Machine-learning integrated multi-domain co-optimization for electrified heavy duty fleets

eTransportation, vol. 27, pp. 100511

Abstract

Driven by global regulations and the urgent need for a sustainable transition to zero-emission fleets in the transport sector, revolutionizing powertrain systems and their respective development processes have become more and more prevalent. Ambitious goals have been established for the latest public-funded research projects, such as ESCALATE (Powering European Union Net Zero Future by Escalating Zero Emission Heavy Duty Vehicles (HDV) and Logistic Intelligence), to increase the efficiency of the powertrain by up to 10% and thus maximize the operational range above 750 km. All of this will be achieved by introducing cost-effective, modular, and scalable electric powertrain components combined with advanced system control algorithms, targeting a broad market coverage with flexible vehicle architectures. In this context, the paper presents a completely virtual frontloading strategy to create a modular and highly integrated e-Axle system, leveraging a dual permanent magnet synchronous machine configuration to improve multiple performance indicators. These are the performance output, in terms of power and torque, system efficiency, and noise-vibration-harshness (NVH) criteria. To allow for an holistic system parametrization, a combined electric machine and transmission synthesis, using an active learning-based, multi-layer nested optimization approach together with a model predictive control strategy for motion and thermal domain has been employed. This development framework is integrating electric machine dimensions and transmission gear ratios as design parameters, as well as thermal actuation and torque as control parameters, to ensure a system right-sizing in a given use-case environment. By including monetary considerations with genetic algorithms, an extension for a powertrain family identification to a complete HDV fleet is facilitated. To demonstrate the feasibility of this framework, a concept assessment and validation has been carried out. The key achievements include a close matching of the defined KPIs, namely the peak wheel torque of 56150 Nm and continuous power of 381 kW - about 2%, respectively 0.2% above the target - and an enhanced peak power capability of 536 kW. In terms of energy efficiency, the multi-stage gear boxes support a well optimized operation in the VECTO long haul cycle, indicating a 40-ton vehicle energy consumption of around 109.7 kWh per 100 km, while the 76-ton variant consumes approximately 204.6 kWh per 100 km. Further the predictive cruise control strategy led to a consumption reduction of about 2.6–3.4%. • Machine-learning integrated co-optimization for heavy-duty fleet e-axles. • Dual-PMSM and 2-speed gearbox synthesis with predictive control. • Achieved 56,150 Nm peak torque and 381 kW continuous power in validation. • Energy consumption of 106.8 kWh/100 km (40t) and 204.6 kWh/100 km (76t). • Predictive cruise control reduced consumption by additional 2.6–3.4% • Framework enables modular, scalable powertrain family design for HDVs.

Authors 8

1. Christoph Wellmann Aachen Faculty of Mechanical Engineering

   RWTH Aachen University

   Affiliation as printed

   Teaching and Research Area Mechatronics in Mobile Propulsion, Faculty of Mechanical Engineering, RWTH Aachen University, Forckenbeckstraße 4, 52074 Aachen, Germany

2. Pekka Rahkola

   VTT Technical Research Centre of Finland

   Affiliation as printed

   VTT Technical Research Centre of Finland Ltd, P.O. Box 1000, FI-02044 VTT, Finland

3. Mikko H. Pihlatie

   VTT Technical Research Centre of Finland

   Affiliation as printed

   VTT Technical Research Centre of Finland Ltd, P.O. Box 1000, FI-02044 VTT, Finland

4. Abdul Rahman Khaleel Aachen Faculty of Mechanical Engineering

   RWTH Aachen University

   Affiliation as printed

   Teaching and Research Area Mechatronics in Mobile Propulsion, Faculty of Mechanical Engineering, RWTH Aachen University, Forckenbeckstraße 4, 52074 Aachen, Germany

5. Christopher Marx Aachen FEV Europe GmbH

   FEV (Germany)

   Affiliation as printed

   FEV Europe GmbH, Neuenhofstraße 181, 52078 Aachen, Germany

6. Akshay Sharma Aachen FEV Europe GmbH

   FEV (Germany)

   Affiliation as printed

   FEV Europe GmbH, Neuenhofstraße 181, 52078 Aachen, Germany

7. Markus Eisenbarth Aachen Faculty of Mechanical Engineering

   RWTH Aachen University

   Affiliation as printed

   Teaching and Research Area Mechatronics in Mobile Propulsion, Faculty of Mechanical Engineering, RWTH Aachen University, Forckenbeckstraße 4, 52074 Aachen, Germany

8. Jakob Andert corresponding Aachen Faculty of Mechanical Engineering

   RWTH Aachen University

   Affiliation as printed

   Teaching and Research Area Mechatronics in Mobile Propulsion, Faculty of Mechanical Engineering, RWTH Aachen University, Forckenbeckstraße 4, 52074 Aachen, Germany