Research on Electric Vehicle Powertrain Systems Based on Digital Twin Technology

Abstract

As a critical component of electric vehicles, the powertrain has a significant impact on the overall performance of vehicles. In addressing the challenge of lengthy testing cycles, this study develops a para model of the powertrain, utilizing digital twin (DT) technology, thereby establishing a framework for simulation testing of multi-controller intermodulation. We establish functional definition coverage testing by designing specific functional requirement use cases, and we validate the failure mechanism via fault injection use cases. The results indicate that the DT testing platform can effectively simulate the operational interactions among various controllers within the powertrain system. In comparison to traditional field testing, the digital twin-based testing methodology offers enhanced operational efficiency and allows for the examination of testing conditions that are impractical to implement in real vehicles, particularly in the context of fault injection testing, thus facilitating the early detection of potential safety risks within the system. The advancement of this technical solution holds significant practical implications for the future mass production and development of electric vehicles.

1. Introduction

In the current global process of the technological revolution and industrial transformation, automobiles, deeply integrated with energy, transportation, and communication technologies, have garnered significant attention from industry, academia, and research institutions. With the trends of electrification and intelligent networking driving the automotive industry, new energy vehicles (NEVs) integrate cutting-edge technologies, such as new energy sources, innovative materials, big data, and artificial intelligence (AI). This transformation from traditional transportation tools to mobile intelligent terminals, energy storage units, and digital spaces supports the construction of a clean and beautiful world and the building of a community with a shared future for mankind [1]. In recent years, supported by strategic planning and policies globally, NEVs have emerged as the primary direction for the transformation and development of the global automotive industry and as a key driver for sustainable economic growth. Developing NEVs is an essential path for countries to become automotive powerhouses and serves as a critical strategy to address global climate change and promote green energy development [2,3].

From a developmental perspective, pure electric vehicles (PEVs), range-extended hybrid electric vehicles (REHEVs), plug-in hybrid electric vehicles (PHEVs), and fuel cell electric vehicles (FCEVs) are the main future product routes, offering stronger competitive and attractive advantages compared to traditional vehicles. Among these, PEVs have become the focus of research and industrial layout investments for automotive enterprises due to their relatively straightforward technological implementation, high user acceptance, and strong governmental support [4,5].

During the marketization of electric vehicles, the powertrain system, as a core component, profoundly influences the vehicle’s overall performance and user experience. The current industry commonly employs three pillars for powertrain system functional testing and quality assurance: simulation testing, closed-track testing, and real-vehicle testing [6]. Among them, closed-track testing offers controlled environments and high repeatability but is limited to a controlled environment, which cannot fully cover the working conditions of the product. The real-vehicle testing, conducted on open roads, benefits from authentic traffic conditions but poses risks of traffic interference and extended test cycles. Simulation testing, using computer models and virtual environments, assesses and verifies the performance and safety of power systems at a lower cost and with flexible testing scenarios, replicating specific operational conditions and providing abundant test data.

Digital twin (DT) technology involves creating a virtual simulation model corresponding to a real-world object, updated and optimized in real time through data feedback. Widely applied in automotive manufacturing, aerospace, and healthcare, DT enables simulation mapping of physical components and constructed models in virtual spaces during simulation tests, reflecting the actual operational status of physical equipment and enhancing development and testing efficiency [7,8]. Particularly in the electrical vehicle sector, to adapt to the development concept of a short iteration cycle of electric vehicle models, frequent product updates, and fast time-to-market, the digital twin simulation test before the prototype trial production, as an indispensable design method, can provide maximum help for the enterprise to develop their products as quickly as possible, and to save the cost of expenses, which is vigorously promoted by the automotive industry [9]. Therefore, the application of DT technology has gained promotion within the field of automotive testing.

In the process of landing the application of digital twin simulation test technology, different departments adopt different focus points and technical paths. Regarding technical path realization, Zhao Shulian et al., from the China Automotive Research Institute, proposed a digital twin-based test and evaluation method for intelligent vehicles. They combined real test vehicles and simulation test tools to build a digital twin autonomous driving test platform to achieve the verification of algorithms and to achieve the effective mapping and combination of vehicles traveling in the real test site within the virtual test environment [10]. Based on the designed data acquisition platform, Zhang Shiyuan of Chang’an University conducted research in three parts: system hardware deployment and connection, sensor data acquisition and processing, and system communication connection and data transmission. They also constructed a digital twin test landscape and proved through experiments that the trajectory prediction method can control the positional error between the digital twin vehicle and the real vehicle to the centimeter level [11].

In the context of simulation testing, software testing can validate dynamic functionalities by executing software within a virtual environment. Dr. Olena Ivanova’s team proposed a Model-in-the-Loop (MiL) design criterion suitable for system behavior models and safety-critical constraints, taking into account safety objectives and system dynamics characteristics. This approach aims to reduce the coordination efforts needed in distributed development [12]. Additionally, Zhentian Qian and colleagues established a System-in-the-Loop (SiL) simulation system using MATLAB 2022b software and code compilers, conducting simulation experiments on circuit control systems to shorten development cycles [13]. However, both SiL and MiL testing cannot fully replicate the complex dynamic behaviors of systems exhibited by hardware during actual operation, leading to challenges in identifying issues related to hardware–software coupling. To further validate the system response behavior that combines real hardware with simulation environments, Jiangnan Yu and his team from Wuhan University of Technology developed a pure electric vehicle model and a rapid prototype controller for Hardware-in-the-Loop (HiL) testing. They conducted closed-loop simulation tests between the controlled vehicle model and the control strategy, verifying the effectiveness of HiL simulation testing from a functional perspective [14]. Meanwhile, Jiangbo Liu and colleagues at JAC Motors set up a multi-ECU (Electronic Control Unit) collaborative debugging platform for HiL testing, assessing the rationality of design matching among different ECUs, to achieve reliability studies on the overall electronic and electrical architecture of the vehicle [15]. However, the accuracy of HiL testing depends on the precision of the test environment model. Achieving model quality necessitates the construction of various types of simulation signals with different timing, which incurs significant development and maintenance costs.

Despite the breakthroughs achieved by the aforementioned scholars in the implementation of digital twin technologies and the simulation testing of automotive products, their findings often focus primarily on the evaluation of individual performance metrics. This tendency overlooks the hierarchical and complex nature of the overall system, resulting in experimental outcomes that may not comprehensively reflect the performance of the power system under diverse operating conditions. Therefore, there is an urgent need to develop a testing methodology for power systems that is characterized by real-time capabilities and high precision. This methodology should accurately reflect the status and behavior of actual systems, providing immediate feedback on changes in environmental and operational conditions. Such an approach would enable comprehensive testing of the fundamental functions and fault-handling mechanisms of electric vehicle power systems, improving the scientific and practical aspects of product development. Strategically, this development would help alleviate concerns regarding lengthy research and development cycles and uncontrolled development risks within vehicle R&D organizations, thereby providing a foundation for the optimization and upgrading of electric vehicles.

Therefore, this study proposes the Sandian testing scheme based on a digital twin framework, focusing on the testing of the powertrain components, including the motor, battery, and electronic control systems. The primary advantage of this testing approach lies in its ability to effectively mitigate personnel injury during the experimental process while significantly enhancing testing efficiency. The key benefits of implementing this scheme include the replacement of potentially hazardous signals during actual driving with simulated signals generated by the digital twin, thereby avoiding unsafe conditions for testers during the functional verification phase of the electric vehicle’s powertrain. Another advantage of the three-in-one system testing is its capability to substantially reduce the experimental costs associated with the electric vehicle development phase. Compared to manual road testing with real vehicles, testing on a powertrain DT platform allows for significant flexibility in selecting controlled experimental conditions and parameter settings according to testing requirements, thereby alleviating the risks associated with lengthy research and development testing cycles. Additionally, the Sandian testing system supports the reproduction of fault conditions, enabling the focused recreation of critical test scenarios, which facilitates data collection and problem identification by the experimental personnel. In summary, the proposed testing based on DT technology effectively controls testing risks and optimizes testing costs while ensuring reliable results.

Considering that testing with a single critical component controller does not fully reflect the signal logic interactions of the powertrain during actual operations, and recognizing that real-vehicle testing fails to accommodate rapid parameter design and safe implementation of extreme operating conditions, we propose a digital twin simulation testing strategy based on a multi-controller framework. This study focuses on the functional testing of electric vehicle powertrains and presents a detailed technical scheme for the multi-controller simulation platform through the analysis of the characteristics of digital twin testing methods. Building upon the functional definitions provided by the vehicle development phase, testing use cases are developed based on the characteristics of the simulation platform. This enables functional testing under common operating conditions and powertrain testing under fault injection scenarios with extreme parameter conditions. Through data analysis and criterion evaluation, this approach offers substantial support for the development of functional strategies and the enhancement of fault-handling mechanisms.

The organization of this article is as follows: Section 1 introduces the discussion topic. Section 2 presents the Digital Twin Powertrain System, providing a methodological framework and guidance for the research. Section 3 outlines the technical approach of the digital twin test bench, including important information about digital twin architecture, hardware configurations, and software implementations. Section 4 details the specific plans and implementation methods for the construction of the experimental test bench, covering hardware connections and digital twin software modeling. Section 5 compiles test cases for the digital twin experiments based on the characteristics of this project, categorizing them into functional tests and fault injection tests. Section 6 focuses on testing the prepared test cases and analyzing and discussing the results of the digital twin tests. Section 7 offers a conclusion, extending and projecting the findings of the research.

2. Digital Twin Powertrain System Framework

In relation to the reference architecture for digital twins, the team led by Tao Fei at Beihang University proposed a five-dimensional model, which specifically includes Physical Equipment, Virtual Equipment, Connection, Digital Twin Data, and Services. These five modules facilitate the integration of cyber-physical systems, the fusion of cyber-physical data, bidirectional interaction between the virtual and physical realms, and data services [16]. In the context of electric vehicle powertrains, the Vehicle Controller Unit (VCU, Lianhe Dianzi, Wuhu, China), Motor Controller Unit (MCU, Shanghai Dianqudong, Shanghai, China), and Battery Management Unit (BMS, Guochuang Xinneng, Hefei, China) are essential critical components, and they become focal points for simulation testing research [17]. Accordingly, this study builds upon the aforementioned five-dimensional model to propose a “Twin-Reality-User Domain” framework for testing critical components within the powertrain. This approach supports the simulation of the physical processes and system behaviors of the VCU, BMS, and MCU under real vehicle operating conditions by establishing mathematical models within a computer environment. This facilitates testing, research, and optimization from the source level.

As shown in Figure 1, the DT powertrain system consists of three components: the twin module centered on the simulation platform, which primarily includes the virtual test environment and testing scenarios; the physical entity module centered on the tested objects and real testing environment, which mainly encompasses industrial computers and key data processing devices; and the user domain module focused on driving and monitoring, with the primary goal of evaluating the test result data.

During system operation, the simulation platform generates the necessary simulation signals to maintain the proper functioning of the tested controllers, including bus communications (CAN), hardwired signals, and Ethernet signals, collectively constructing a virtual digital vehicle. The virtually constructed vehicle transmits target test information to the powertrain’s critical component controllers via the industrial computer and data injection devices. Subsequently, the physical controllers respond to the received operating condition and environmental parameter information, while outputting typical powertrain parameters (such as voltage, current, torque, and speed) to the virtual digital vehicle model, effectively mapping the operational state of the controller within the simulation environment, thus realizing the simulation of the “digital-physical” system.

During this process, the user interface outputs simulated driver-triggered signals (such as throttle and brake signals) to the closed-loop simulation model. The real-time transmission method based on the TCP/UDP protocol synchronizes the vehicle’s operating status between the virtual world and the real environment. At the same time, adjustments are made in response to data monitored by dynamic measurements from monitoring devices, ensuring the stable closed-loop operation of the entire digital twin testing platform.

The digital twin technology involves the digitization of physical characteristics, operating states, and environmental information of tangible objects. Its fundamental concept is to collect data from the physical world, establish corresponding digital models, and seamlessly connect these models with the physical objects in virtual space [18,19]. The proposed “Twin-Reality-User Domain” system architecture demonstrated in this paper exhibits high-fidelity characteristics, ensuring precise and reliable simulations and providing a robust platform for the testing and verification of electric vehicle powertrains.

3. Technical Scheme of Digital Twin Test Bench

In addressing project requirements, this section focuses on the practical implementation of a digital twin simulation platform based on the powertrain digital twin system architecture. The simulation platform consists of three main components: hardware, software modules, and the tested objects [20]. Together, these components simulate the relevant parameters of the critical controllers within the powertrain, laying the groundwork for the simulation testing of these controllers.

This study examines the test subjects, specifically the combination of the Vehicle Control Unit (VCU), Battery Management System (BMS), and Motor Control Unit (MCU), within a three-electric system. A multi-digital twin simulation approach is employed to facilitate signal interactions among essential controllers within the power system. Below is a schematic for the digital twin platform.

As depicted in the Figure 2, the initial phase of the process includes system re-inspection and the identification of the master cabinet. The re-inspection ensures the overall cabinet comprises an industrial computer, a monitoring and data acquisition system, and a real-time control system. A specific hardware device is designated as the master cabinet; this device is typically selected based on its ability to perform the most complex computations required for the dynamic model [21]. If a cabinet employs multi-core processing during debugging, that section is identified as the master cabinet. The communication network among these cabinets is subsequently planned. Serial connections are utilized, allowing a maximum of one cabinet to communicate with two others. For example, in a setup with three or more cabinets (Cabinet 1—Cabinet 2—Cabinet 3—Cabinet X), communication between Cabinet 3 and Cabinet 1 must occur through Cabinet 2’s forwarding. To minimize unnecessary delays caused by signal forwarding among cabinets, a section with relatively high signal interactions is selected as the master cabinet.

Secondly, during the hardware configuration stage, pre-installed DSPACE (A German Automobile parts manufacturer) real-time processors, IO boards, bus communication boards, and Field-Programmable Gate Array (FPGA) boards are prioritized for selection as the real-time control system, grounded in the electrical characteristics of the test subjects. Corresponding digital-to-analog acquisition modules and CAN communication modules are matched to enable authentic data communication with the controller. In turn, wire harnesses linking the platform and controller are constructed according to assigned electrical channels, ensuring that the wire gauge of the finished harnesses meets testing specifications.

Thirdly, for the upper computer software platform, dynamic simulation software (ASM 2022) is used to establish the vehicle dynamics model, while MATLAB 2022b is employed to create the power system simulation platform. The configuration software configures the overall experimental setup, and the Control Desk 2021 is utilized to develop the testing interface and edit typical environmental parameters, with key signals designated for subsequent activation. The schematic diagram of DT bench technology scheme is showed in the Figure 3.

Additionally, the main tasks involved in the debugging of the simulation test platform include the open-loop verification of input and out interface resources based on the defined signal list and the selection of certain controller feedback functions for closed-loop operation. Ultimately, the digital test platform system is debugged into a working state, which provides a foundation for subsequent testing.

Finally, digital twin simulation tests are conducted, with relevant signals recorded throughout the process. Taking the basic signals of the vehicle as an example, the technical solution implemented by the platform follows this path: The simulation platform converts the corrected electrical signal information into throttle position or brake switch signals to control the digital twin vehicle. This is accomplished through a joint simulation platform built with MATLAB/Simulink for the simulation and verification of safety control strategies. This technical solution supports the injection of signals from the source end to ensure the high precision and reliability of the simulation, thereby maximizing the replication of real vehicle operating conditions.

4. Testing Platform for Sandian Digital Twin System

The digital twin system testing platform consists of two main components: hardware and software [22]. This section provides a detailed discussion of the structure and construction logic of each component based on the specific requirements of the Sandian DT system project.

4.1. Hardware Configuration

The focal point of the Sandian testing system platform is the connection between the tested components and the cabinet, particularly the data connections between the actual controllers and their corresponding digital twins [23]. Given the characteristics of multi-cabinet coordination, the wiring arrangement and linkage for the critical components of the powertrain system—VCU, MCU, and BMS—are outlined in

Table 1. Table of linking methods for each controller of the powertrain.

	VCU	BMS	MCU
VCU	/	Er	Pr
BMS	Er	/	VCUt
MCU	Pr	VCUt	/

.

In the table, Er indicates that the two controllers in the E-CAN lines are connected as actual lines, Pr indicates that the two controllers in the P-CAN lines are connected as actual lines, and VCUt indicates that the two controllers are connected by forwarding signals through virtualized VCUs.

Considering that the VCU serves as the central control unit for the vehicle power system, this project adopts a configuration where Module 1 links to the VCU as the main cabinet, while Cabinets 2 and 3 are connected to the BMS (including both primary and secondary components) and MCU, respectively. Signal interaction among the three cabinets is facilitated through optical cables and Ethernet, while signal transmission and communication between the cabinets and the control unit under test are conducted via the CAN bus. The schematic diagram of the Sandian testing hardware connections is illustrated in Figure 4 below.

4.2. Software Modeling

The software component of the digital twin testing platform is primarily responsible for translating the computational logic and intrinsic relationships between the signals of various controllers into unified mathematical expression models for testing purposes [24]. This section discusses typical components of the three-in-one system: the vehicle para model, motor para model, and battery para model. According to the project database, the vehicle characteristics are summarized in

Table 2. Table of parameter list of the electric vehicle.

Parameters	Unit	Value
Vehicle mass	(kg)	1842
Full load mass	(kg)	2217
Drivetrain efficiency	/	0.95
Rolling resistance coefficient		0.007
wind resistance	/	0.32
windward side (of an area)	(m2)	2.698
rolling radius	(mm)	360
Motor rated power	(KW)	70
Peak motor torque	(NM)	285
Battery pack available voltage range	(V)	228~416

.

The “/” means a parameter is represented as a pure numerical value, without units.

4.2.1. Vehicle Longitudinal Dynamics DT Model

The vehicle longitudinal dynamics model plays a critical role in automotive engineering by simulating the dynamic behavior of the vehicle under various driving conditions, particularly during acceleration and braking, thus assisting engineers in analyzing vehicle handling and safety [25]. According to automotive theory:

$${\mathrm{F}}_{\mathrm{t}}={\mathrm{F}}_{\mathrm{f}}+{\mathrm{F}}_{\mathrm{w}}+{\mathrm{F}}_{\mathrm{i}}+{\mathrm{F}}_{\mathrm{a}}$$

(1)

where F_t is the driving force of the whole vehicle; F_f is the rolling friction resistance; F_w is the air resistance when the vehicle is traveling; F_i is the ramp resistance of the vehicle on an uneven road; and F_a is the resistance due to acceleration of the vehicle.

$${\mathrm{F}}_{\mathrm{t}}=\mathrm{m}\mathrm{g}\mathrm{f}\cos \beta +0.5\mathsf{\rho }\mathrm{A}{\mathrm{C}}_{\mathrm{d}}{\mathrm{V}}^2+\mathrm{m}\mathrm{g}\mathrm{f}\sin \beta +\mathsf{\delta }\mathrm{m}{\displaystyle \frac{\mathrm{d}\mathrm{v}}{\mathrm{d}\mathrm{t}}}$$

(2)

where $\beta$ is the road slope; m is the vehicle’s overall mass; g is the gravitational acceleration; ρ is the air density at room temperature; and δ is the vehicle’s rotational mass factor.

4.2.2. Vehicle Controller Unit DT Model

In the Sandian simulation testing, the vehicle controller VCU simulation model typically consists of several key components, each playing a crucial role in ensuring the system meets the expected performance and functionality [26]. The main functions of the VCU model components are as follows:

Sensor Simulation Module: This module is responsible for simulating the input signals from various sensors within the vehicle system, such as speed and steering sensors. It provides real-time environmental data and ensures coordination between the sensors and the VCU. Actuator Simulation Module: This module simulates the actuators of the vehicle’s powertrain, such as the engine and brake actuators. It receives control commands from the VCU and simulates the actual behavior of these actuators. Communication Interface Module: This module simulates the communication process between the VCU and other vehicle systems, including the reception of sensor data and the transmission of control commands. It ensures normal communication between the VCU and the entire vehicle system.

The VCU digital twin simulation model is constructed as illustrated in Figure 5.

By coordinating these components and implementing their functions, the VCU simulation model effectively simulates the behavior and performance of the entire vehicle system during DT testing, thereby supporting system integration validation. To be specific, the VCU DT model encompasses not only the VCU hardware itself (VCU body) but also the MDL module that simulates the operation of the VCU (para model), the power module that controls the power supply to the cabinet, the Bus module that facilitates communication between power systems, and the HiL-Communication module that manages signal forwarding between different cabinets.

4.2.3. Motor Controller Unit DT Model

The main components and their functions in the MCU simulation model in the Sandian test are a communication interface module to achieve data transmission and command exchange; a signal acquisition module to simulate the signal acquisition function in the MCU of the motor controller, including receiving feedback signals from sensors, such as speed, temperature, etc.; a control algorithm module to achieve speed control, torque control, position control, etc.; a PWM generator module to simulate PWM signal generation function for driving power semiconductor devices in the motor controller; a protection logic module to achieve overcurrent protection, voltage protection, temperature protection, etc., of the controller to avoid damage to the controller. The PWM generation module simulates the PWM signal generation function, which is used to drive the power semiconductor devices in the motor controller; the protection logic module realizes the controller’s overcurrent protection, overvoltage protection, over-temperature protection, etc., to prevent the controller from suffering potential damage [27].

The primary parameters of the motor model are determined by the slope of the operating environment, the vehicle’s mass, and the transmission efficiency of the power system [28], as expressed by the following equation:

$${\mathrm{T}}_{\mathrm{m}}={\displaystyle \frac{\mathrm{mgfcos}\left(\mathrm{arctg}\right(\mathrm{grade}/100\left)\right)\mathrm{r}}{{\mathrm{i}}_0\mathsf{\eta }}}+{\displaystyle \frac{\mathrm{mgsin}\left(\mathrm{arctg}\right(\mathrm{grade}/100\left)\right)\mathrm{r}}{{\mathrm{i}}_0\mathsf{\eta }}}$$

(3)

where T_m is the peak torque of the motor; grade is the climbing parameter set for the vehicle; i₀ is the powertrain transmission ratio; $\mathsf{\eta }$ is the system efficiency.

The voltage equation of the permanent magnet motor based on rotor field orientation is obtained as follows:

$$u_d=R_s{i}_d+L_d{\displaystyle \frac{d{i}_d}{dt}}-{\omega }_r{L}_q{i}_q$$

(4)

$$u_q=R_s{i}_q+L_q{\displaystyle \frac{d{i}_q}{dt}}+{\omega }_r(L_d{i}_d+{\psi }_f)$$

(5)

where R_s, L_d, and L_q are the stator winding resistance, stator d-axis inductance, and stator q-axis inductance. Ψ_f is the magnetic chain of permanent magnets. ω_r is the rotor electric angular velocity. i_d and i_q are the d-axis and q-axis currents. u_d and u_q are the d-axis and q-axis voltages, respectively.

The motor torque T_e formula is given by:

$${\mathrm{T}}_{\mathrm{e}}=\mathrm{p}[{\Psi }_f+(L_d-L_q)i_d]i_{\mathrm{q}}$$

(6)

where p is the number of motor pole pairs.

Accordingly, the MCU digital twin simulation model is built as shown below Figure 6. The MCU DT model not only includes the motor hardware itself (motor) but also incorporates a mechanical module (mechanic) that reflects the mechanical connection principles. Additionally, it features an inverter module (inverter) responsible for converting direct current (DC) to alternating current (AC) and a DC-link module that manages and applies the DC provided by the battery. The environment control module simulates the parameters of the MCU operating environment, while the Proc-IO module supplies power to the cabinet. Furthermore, the Bus module facilitates communication within the power system, and the HiL_Communication module handles signal forwarding between different cabinets.

4.2.4. Battery Management System DT Model

The main components and their functions of the battery management system simulation model in the DT test are listed: the communication interface module is responsible for simulating the communication process between the battery controller BMS and other vehicle systems; the data acquisition module realizes the collection and monitoring of battery parameters, such as voltage, current, temperature, etc.; the state estimation module simulates the state estimation algorithms in the battery controller BMS, which is used to estimate the battery’s residual capacity and health status; the protection control module simulates the protection functions in the battery controller BMS, including overcapacity protection, over-discharge protection, short-term protection, etc., information; the protection control module simulates the protection functions in the battery controller BMS, including overcharge protection, over-discharge protection, short circuit protection, etc. [29].

The model of the battery addresses the mathematical relationship between the state of charge (SOC) of the battery and the characteristic parameters of the battery [30]. This is shown in the following equation:

$$\mathrm{SOC}={\mathrm{SOC}}_{\mathrm{int}}-{\displaystyle \frac{1}{\mathrm{Q}}}\eta {\displaystyle \int {\mathrm{I}}_{\mathrm{t}}}\mathrm{dt}$$

(7)

$${\mathrm{I}}_{\mathrm{t}}={\displaystyle \frac{({\mathrm{U}}_{\mathrm{t}}-\sqrt{{{\mathrm{U}}_{\mathrm{t}}}^2-4{\mathrm{R}}_{\mathrm{t}}{\mathrm{P}}_{\mathrm{bat}}})}{2{\mathrm{R}}_{\mathrm{t}}}}$$

(8)

In these equations, SOC_int is the initial state of charge; η is the Coulombic efficiency of the battery; Q is the capacity of the battery. U_t is the open-circuit voltage of the battery; R_t is the internal resistance of the battery; and P_bat is the output power of the battery.

The BMS digital twin simulation model is constructed as illustrated in the following Figure 7. The BMS DT model includes not only the BMS hardware components, specifically the BMS-Body-In and BMS-Body-Out modules, but also comprises the MDL module that simulates BMS operation, the power module that controls the cabinet power supply, the Bus module responsible for communication between the power system, and the HiL_Communication module that facilitates signal forwarding between different cabinets.

5. Test Case Design of the Sandian Platform

During the development of test cases for the digital twin simulation platform of the powertrain system, functional logic testing of typical scenarios and fault injection testing under severe conditions are designed to ensure that the parameters of the test cases align more closely with the predefined requirements established for the powertrain system. This design facilitates a scientific evaluation of the system’s functionality [31]. Therefore, in response to the functional definitions provided by the R&D department, this section undertakes the development of functional requirement use cases based on relevant standards. Additionally, leveraging the technical knowledge and project experience of the project team members, targeted fault injection test cases are designed for both functional scenarios and extreme corner cases of the powertrain system.

5.1. Functional Requirement Test Case Design

According to industry testing guidelines, each function must have corresponding individual measurable requirements, with each measurable requirement linked to one or more specific test cases [32]. Consequently, the functional definitions of the powertrain system are decomposed into measurable requirements, enabling the subsequent design of test cases using clear pass/fail criteria to verify the predefined requirements. Based on the project characteristics, the following methods are employed to design derived test cases for the powertrain system based on functional requirements:

①

Equivalent partitioning

Test cases are crafted to cover all acceptable valid data and reject invalid partitions [33]. For valid and invalid data, equivalence classes are identified, considering specific parameters such as output values, internal values, and time-related values.

②

Boundary value analysis

Boundary value analysis complements equivalence partitioning analysis. The behavior of the powertrain system at the boundary values of specific parameters is often more unpredictable than its behavior within the partitions, making it more likely to reveal undiscovered defects. Therefore, this technique is typically used as an extension of equivalence partitioning or other black-box testing design techniques [34].

③

Assumption error

The assumed error analysis method employs structured thinking by listing potential defects or failures and designing tests aimed at triggering these defects or failures. This list is developed based on the experiences of the testers and their colleagues, along with the collection of data regarding typical defects and failure scenarios. It embodies the collective knowledge and experience of the team.

5.2. Fault Injection Test Case Design

Fault injection represents a reliability validation technique that involves injecting faults into a system through controlled experiments to observe its behavioral responses and coping mechanisms [35]. During system development, the dependency of software components on other system entities regarding functional failures is analyzed using supervised execution in conjunction with simulation technology. The application of fault injection techniques in powertrain simulation testing enables testers to observe the safety response mechanisms of signal interactions between various components under fault-injection conditions within the simulation environment, facilitating an assessment of the powertrain system’s reliability.

The main significance of the design of functional requirements and fault injection use cases is to study the response of important functions of the powertrain in a digital twin test environment. By analyzing the expected goals, the project team designed the use cases to cover the two major categories of component system fault injection testing and vehicle functional testing based on the separate open- and closed-loop debugging of VCU, MCU, and BMS and the digital twin testing strategy of multi-controller intermodulation. A total of 673 functional requirement test cases and 419 system fault injection test cases were designed, with a consideration of ethical responsibilities. The test cases are executed for subsequent use cases to achieve full coverage of the test, which makes the system test focus evolve from purely functional to robustness and facilitates the execution and optimization of subsequent system regression tests.

6. Test Results

The hardwired configuration and connections of each controller are established according to the topology diagram defined by the actual vehicle model, while the signal transmission channels of the test cabinet and software modules are debugged. The digital twin testing platform is constructed as shown Figure 8 below.

During the testing of the three-electric system, strict environmental conditions are implemented. The temperature is maintained within a constant range of 10–45 °C, and humidity is controlled between 40 and 60%. Additionally, anti-static measures are taken to protect circuit boards and components from electrostatic interference. The testing environment is also equipped with specialized power regulation devices and signal generators to ensure stable power supply and signal input conditions during hardware testing. By rigorously controlling environmental conditions, external factors affecting the test results are effectively eliminated, ensuring the accuracy and reliability of the data collected.

According to the test requirements of the three power systems, we designed 673 test cases in the functional requirement category, containing 13 items, up and down function, slow-charging function, fast-charging function, driving mode, inverter function, energy recovery, creep braking, intelligent replenishment, gear EPB, slow-charging lockout, water cooling of batteries, battery thermal loss of control, and drive driving function, as well as 419 test cases of system fault injection, containing 4 items of VCU, BMS, MCU and system class 4 items.

This section selects the power up/down and driving condition validations from the functional requirement test cases to assess the effectiveness of the digital twin three-electric system testing. Furthermore, typical voltage-related fault injection scenarios are selected to discuss the comparative advantages of three-electric simulation testing versus actual vehicle testing.

6.1. Functional Requirements Test Result

In the 13 sub-items of the real-vehicle functional tests, adherence to the predetermined development flowchart is maintained. The flowchart in the Figure 9 primarily delineates the process mechanism for signal detection and fault status determination during the reception of driver operation instructions following system initialization. This process establishes a foundation for the functional validation of subsequent digital twin testing.

Based on the principles of multi-controller participation, typical power up/down conditions of the powertrain system are tested using the digital twin platform.

As shown in the Figure 10, during the power-up process of the powertrain system, the status flag signals of the BMS and MCU, specifically BMS_BattReadySts and RMCUWorkingModSts, transition from their initial default states (1 and 5, respectively) to the high-voltage standby state (0) at the 3.5 s mark. After 1.8 s in the high-voltage standby state, the BMS indicates the statuses of the main negative relay (BMS_BattRelay_Neg) and main positive relay (BMS_BattRelay_Pos), which switch from open (0) to closed (1), marking the completion of high-voltage power supply to the system. Once the VCU confirms the high-voltage power supply completion feedback from the BMS and MCU, it issues a command (VCU_HV_PowerOnSts = 1), which changes the motor status signal (RMCUWorkingModSts) from the high-voltage standby state (value 1) to the operational state (value 2). Subsequently, provided that there are no faults in the feedback from each controller, the VCU transitions the powertrain system to the ready state (VCU_PowertrainReady = 1).

At the power-down moment, the powertrain system first changes from the ready state to power down (VCU_PowertrainReady = 0). Then, at the 22 s mark, the VCU issues a high-voltage power down command (VCU_HV_PowerOnSts = 0), resulting in the disconnection of the main positive and negative relays of the battery. The status flag signals of the BMS and MCU (BMS_BattReadySts and RMCUWorkingModSts) revert to their initial default states. The three-electric testing platform successfully completes the testing of power up/down conditions.

The above diagram illustrates the signal timing for the driving and braking conditions of the powertrain system. As seen in the Figure 11, at the 11 s mark, when the VCU-related signals indicate that the accelerator pedal position (VCU_AclPedalPos = 0) and brake pedal switch (VCU_BrkPedalSts = 0) are released, the vehicle begins to accelerate due to the creep functionality, causing the vehicle speed signal (VCU_VehSpd) to gradually increase. At the 13 s and 15.5 s marks, the driver operates the accelerator pedal position, increasing it from 0 to 10% and then to 20%. Consequently, the torque feedback from the MCU (RMCU_ActTorq) begins to increase non-linearly, with the BMS working current (BMS_HV_BattCurr) peaking at 60 A.

At the 27.8 s mark in the figure, the driver depresses the brake pedal, resulting in (VCU_BrkPedalSts = 1). The vehicle speed signal (VCU_VehSpd) decreases from 42 km/h as the system initiates energy recovery, with the MCU feedback torque (RMCU_ActTorq) becoming negative and outputting negative torque. The BMS working current (BMS_HV_BattCurr) also becomes negative, with a peak negative value of −46 A, ultimately reducing the vehicle speed to zero. Thus, the testing of the driving and braking conditions of the three-electric system is completed.

In summary, the recorded parameters from the above tests indicate that the functional logic processes of the digital twin simulation technology and the actual vehicle testing align effectively, demonstrating the validity of the digital twin model. Multiple condition parameters are set as required, enhancing the efficiency of functional testing.

6.2. Fault Injection Test Result

To achieve comprehensive testing of the powertrain system within the scope of fault injection, it is necessary to inject signals that surpass predefined system setpoints. By observing the responses of various controllers to these disturbances, the fault response mechanisms of the powertrain system can be evaluated. However, during actual vehicle testing, the voltage levels at which the drive motor and battery pack operate normally range from 320 to 425 V, significantly exceeding the human safety voltage range of 0 to 36 V. Executing such test cases on actual vehicles poses uncertainties, such as high voltage leakage, which could endanger the safety of testing personnel.

As a result, this section employs a digital twin testing platform to execute the test cases, injecting over-voltage faults into the motor bus voltage during the execution of environmental test cases. The variations in the test result signals are illustrated in the following figure.

As shown in the Figure 12, when the driver presses the accelerator pedal (VCU_AclPedalPos = 10), the vehicle speed signal (VCU_VehSpd) begins to increase from zero. At the 36.6 s mark, the software ControlDesk 2021 is used to forcibly alter the MCU bus voltage signal (RMCU_HV_DClinkVolt) from the normal range of 375 V to an abnormal range of 525 V. This action prompts the MCU to report a level fault at the 39.1 s mark (RMCU_SysFault_Level = 1). Consequently, the VCU_Fault_Level also transitions to 1, causing the system to command the vehicle to decelerate to zero. The BMS working current (BMS_HV_BattCurr) gradually decreases to zero. The changes in signal behavior align with the predefined fault response mechanisms.

6.3. Discussion

The digital twin test bench for powertrain systems simulates conventional driver operations as inputs to observe the key signal variations among various controllers. This approach effectively models the signal interaction processes between different controllers within the powertrain system.

We conduct tests under power-on and power-off conditions, verifying the effectiveness of the digital twin’s power cycling functionality. This is achieved by examining the consistency of the critical temporal changes in the feature representation variables of the battery controller (main positive and main negative relay signals), the motor controller (motor status signal), and the vehicle controller (power-on state signal). Additionally, the hypothesis of the digital twin test’s effectiveness in functional testing is validated through the predefined functions developed during the design process and the trends observed in the driver operation signals during acceleration and braking, alongside the corresponding vehicle status signals (vehicle speed).

The results not only support the hypothesis that digital twin simulation testing is an effective tool in the development of electric vehicles but also validate that virtual testing can accurately reflect real-world conditions across various testing scenarios. Moreover, this research provides the industry with a testing method that enhances development efficiency and safety. Particularly concerning fault injection, the findings confirm that executing use cases in extreme scenarios of digital twin testing can isolate potential personal injury risks to test personnel during real vehicle road tests. This discovery will facilitate ethical considerations in promoting this design methodology.

In summary, the adoption of the digital twin testing method holds the following practical significance in the field of automotive product development:

①

Enhanced Testing Efficiency

Simulation tools can evaluate multiple testing scenarios within a short timeframe without incurring substantial material and labor costs. This high efficiency allows engineers to rapidly adjust experimental parameters and operate independently from external sites and environmental factors, significantly improving testing efficiency.

②

Reduced Project Costs

Digital twin testing is cost-effective compared to traditional testing methods. By eliminating the need for physical prototypes, it can substantially decrease material and production costs. Furthermore, the ability to identify design issues and resolve them during the early stages avoids the high costs associated with later design changes.

③

Promotion of Ethical Considerations

The ethical implications of digital twin testing are significant, particularly in the context of safety verification for power systems. Traditional physical testing often requires test personnel to operate under extreme conditions, potentially exposing them to the risks of accidents and personal injury. Constructing extreme scenario testing within the framework of digital twin methodology alleviates the ethical burdens associated with real-world testing. This approach ensures that the utilized models accurately reflect real-world conditions within a high confidence interval, thereby maintaining user safety and trust alongside technological innovation.

7. Conclusions

7.1. Current Status

• The digital twin testing technology effectively inherits the characteristic of high flexibility in parameter settings from software testing. Due to its capability to efficiently simulate the working environment required by controllers, it is applied in the early development and testing of powertrain systems prior to the prototype manufacturing phase. This technological approach aligns organically with the rapidly evolving demands of the new energy industry and is gradually becoming a focal point for researchers involved in the functional development of electric vehicle powertrains.
• Variations in vehicle models and configurations lead to insufficient coverage in functional testing during powertrain development. This study establishes a digital twin simulation model based on multi-controller joint debugging theory, which accurately simulates the signal interaction processes among different controllers within the powertrain system. It effectively characterizes the intrinsic relationships and computational logic governing signal transmission, thereby becoming a significant driver for product development within enterprises.
• To accommodate the stringent compression of development cycles in powertrain systems, digital twin three-electric simulation testing technology exhibits the same accuracy in functional testing as actual vehicle testing and demonstrates superior advantages in fault injection scenarios. This technology serves as an effective evaluation method for the logical development of critical controller components in electric vehicles, contributing to the ongoing advancement of electric vehicles toward safer and more reliable outcomes.

7.2. Prospects

Future research opportunities include the following:

①

When evaluating the performance of powertrain systems, future studies should prioritize bridging the gaps and differences between laboratory simulations and actual vehicle testing. The first step involves optimizing the scalability of the digital twin para model. For example, while the current model primarily focuses on longitudinal dynamics, it should also consider lateral dynamics and the impacts of other systems on the powertrain.

②

This article presents a research methodology that is applicable to various power system architectures (electrical vehicles/range-extended electrical vehicles/hybrid electrical vehicles/fuel cell electrical vehicles) and different commercial vehicle types, including Passenger Cars and Commercial Vehicles (Buses/Light Trucks/Heavy-Duty Trucks). Additionally, this method is suitable for autonomous vehicles classified at levels L1 to L5. However, in practical implementation, it is necessary to consider the complexity of constructing digital twin models, as well as the compatibility of interfaces between the cabinets and various controllers.

③

One of the typical advantages of the digital twin Sandian testing technology is its efficiency. Future execution methods should evolve toward automation, beginning with the input of requirement documents. The testing platform should automatically provide a comprehensive set of automated testing services, including the automatic generation of test cases, configuration of test programs, execution of tests, recording of test data, and report writing. This evolution will not only significantly enhance testing efficiency but also reduce occurrences of inefficient work practices.