Overview of Digital Twin Platforms for EV Applications

Abstract

Digital twin (DT) technology has been used in a wide range of applications, including electric vehicles. The DT platform provides a virtual representation or advanced simulation of a physical object in real-time. The implementation of DT on various aspects of EVs has recently transpired in different research studies. Generally, DT can emulate the actual vehicle on the road to predict/optimize its performance and improve vehicle safety. Additionally, DT can be used for the optimization of manufacturing processes, real-time condition monitoring (at all levels and in all powertrain components), energy management optimization, repurposing of the components, and even recycling processes. This paper presents an overview of different DT platforms that can be used in EV applications. A deductive comparison between model-based and data-driven DT was performed. EV main systems have been discussed regarding the usable DT platform. DT platforms used in the EV industry were addressed. Finally, the review showed the superiority of data-driven DTs over model-based DTs due to their ability to handle systems with high complexity.

1. Introduction

In recent decades, digital manufacturing has contributed significantly to all industries. The remarkable advances in communication and information technology have gone a long way towards the development of manufacturing [1]. Computer-aided technologies such as computer-aided design (CAD), computer-aided engineering (CAE), computer-aided manufacturing (CAM), finite element analysis (FEA), product data management (PDM), etc., are developing rapidly and play a crucial role in modern industry [2]. Manufacturing, healthcare, and smart city environments have become more able to harness data through advanced analytics and the Internet of Things (IoT) connectivity [3]. In conjunction with data analytics, IoT environments can be used for predictive maintenance, fault detection, and design optimization processes [4].

After Challenge Advisory hosted Michael Grieves' presentation on technology at the University of Michigan in 2002, the concept of the DT gained wider recognition [8]. The term DT refers to the digital representation of a physical object, process, or service that supports decision-making throughout its lifecycle. It is updated from real-time data and uses simulation, machine learning, and reasoning [10]. Three basic pillars make up the DT: the physical entity, the virtual entity, and the data exchange and communication system between them [11]. Recent works have defined DT technology as a five-dimensional structure with separate entities for services and connections [8]. Creating a DT can enhance technology trends, prevent costly failures in physical objects, and improve test processes by using advanced analytics, monitoring, and predictive capabilities.

Historically, automotive and aerospace systems have been developed using empirical engineering practices [13], but now with growing performance requirements, new engineering procedures are required. With the emergence of DT, new testing and development modeling techniques have become available to fulfil new requirements. The EV industry is gaining increased attention nowadays. Pricing and battery capacity pose major challenges to the introduction of EVs on the road. To address these challenges, one way is to optimize the electrical energy consumption of the vehicle and design a supporting architecture to facilitate it. This paper presents a comprehensive overview of different platforms used to develop DTs for EV applications. The general objective of this study is to provide a reference for researchers on this topic.

2. DT Architecture Categorizations

The DT architecture can be divided into two main categories as the following subsections illustrate. The concept of a model-based simulation approach (MBS) refers to a formalized methodology for preparing requirements and designing, analyzing, and verifying complex systems [14]. MBS places models at the center of the system design.

Model-based DT is an advanced form of MBS with increased sensory data and AI supplementary tools. It is clear that the creation of a model-based DT of a system is closely related to modellable physical systems and mostly depends on conventional modeling and simulation platforms. The major drawback of model-based DT is that models cannot handle infinite complexity and typically need to be simplified. Moreover, it has difficulty considering unknown variables and noisy data.

With the substantial increase in process data, conventional model-based methods are unable to describe complex systems' state space. In this way, data-driven modeling technology has become a potential solution for modeling DTs. The data-driven modeling concept is based on analyzing data about a system to find connections between variables (input, internal, and output variables) without explicitly knowing its physical behavior. It is obvious that there is a direct connection between the data-driven DT and the complex systems that contain a huge amount of data. The platforms used for data-driven DT creation are often artificial intelligence and Big Data tools.

3. DT Platforms for EV Applications

EVs are also referred to as battery electric vehicles (BEV), as they use a battery pack to store the electrical energy that powers the electric motor. EV main domains are divided into two categories as follows: a smart vehicle system and a vehicle propulsion drive system.

Emerging technologies in the field of smart vehicle systems have promoted the continuous development of sustainable transport. The smart vehicle system includes advanced driver assistance systems and vehicle health management systems. A DT of an electric passenger bus was developed to emulate the vehicle's performance using the MATLAB/Simulink platform deployed on an NVIDIA processor through CUDA. The Unity game engine was leveraged as a physical system emulator to build a DT of an advanced driver assistance system for connected and automated vehicles.

For the EV battery system, PyBaMM and MATLAB were used to propose hybrid models combining physics-based and data-driven approaches. A cloud BMS was implemented using software programs in Python. A regression model based on sparse-proper generalized decomposition (s-PGD) was incorporated into a DT for real-time simulation and battery evaluation. LIONSIMBA in MATLAB was used for results validation. A semi-analytical DT model was generated to capture thermal behavior using ANSYS.

For the EV electric motor, an intelligent DT model of an EV PMSM was proposed for health monitoring and prognosis using the MATLAB/Simulink platform with ANN and Fuzzy logic tools. MATLAB/Simulink and Unity 3D platforms were used to build a DT of an induction motor for condition monitoring. Siemens' PLM platform was used to build a DT of an airplane electric motor for lifetime fatigue prediction. A DT for an EV propulsion system based on energetic macroscopic representation (EMR) was presented using the LabVIEW platform.

For the traction inverter, a DT of a power converter running on an FPGA was proposed for online diagnostic analysis using the MATLAB platform. A data-driven DT model of a power electronic converter based on a dynamic neural network was developed for condition monitoring using the MATLAB platform. A model-based DT of a power electronic converter was proposed for condition monitoring using the MATLAB/Simulink platform.

Many producers of EVs and their co-systems are using the DT platform for research and development purposes. Some EV manufacturers have established their own DT platforms, while others are in collaboration with global platform developers such as Nvidia Omniverse (BMW), Microsoft Azure (Hyundai), Siemens Xcelerator, and Bosch IoT Suite.

4. Discussion

The first key step of creating a DT for a system is modeling. It is necessary to choose between the two main DT modeling architectures: data-driven and model-based. The selection relies on several factors, including the function performed by the DT, the system parameter availability, and the simplicity or the complexity of the system. The next step of the DT development process is to choose the right development environment (platform). From the perspective of EV applications, EV vehicles were divided into two main domains: the smart vehicle system and the vehicle propulsion drive system. Creating a DT of a smart vehicle system is more achievable based on data-driven techniques. Whilst for EV propulsion systems, a mixture of data-driven, model-based, or hybrid DT architectures have been applied.

The use of platforms such as MATLAB/Simulink, Ansys, LabView, Unity 3D, and other modeling platforms has been effective in creating model-based DTs. While in the case of data-driven DTs, more reliance has been on cloud-based platforms such as Microsoft Azure, AWS, IBM-ELM, or special purposes platforms built by the DT developers based on one of the software development environments, such as Python, C++, R, and others.

5. Conclusions

Recently, DTs have become an emerging paradigm for virtual representations of complex systems along with their underlying components. DTs are composed of three main parts: physical objects, virtual representations, and the communications between them. The virtual part of DT must be developed through a specific environment called the DT platform. Model-based and data-driven are the main categories of DTs. A comparison between the two categories clarified their strengths and weaknesses as well as the prospective applications for both.

This review dealt specifically with DTs for EV applications. EV systems were divided into smart vehicle systems and vehicle propulsion drive systems. The literature addressed the advantages of using data-driven DTs with smart vehicle systems due to the complexity of modeling such systems and also the significant amount of data concerned with it. While in the case of the electric propulsion drive system, there was mixture between the use of model-based DT, data driven DT, or a combination of them both. This paper represents a reference for researchers on the topic of DT for EV applications in order to determine the appropriate DT platform according to the work requirements.

Author Contributions & Funding

Conceptualization, M.I. and V.R.; methodology, M.I.; validation, M.I. and R.G.; formal analysis, M.I.; investigation, R.G.; resources, M.I.; data curation, V.R.; writing—original draft preparation, M.I.; writing—review and editing, M.I.; visualization, V.R.; supervision, M.I.; project administration, M.I.; funding acquisition, M.I. The research has been supported by the Estonian Research Council under grant PSG453 "Digital twin for propulsion drive of an autonomous electric vehicle". The authors declare no conflict of interest.