http://lutpub.lut.fi:80/handle/10024/158303 2026-10-15T23:36:45Z 2026-10-15T23:36:45Z Ocepek, Jakob Nutakor, Charles Sopanen, Jussi http://lutpub.lut.fi:80/handle/10024/173097 2026-10-15T10:30:13Z 2026-06-23T00:00:00Z

Localised Heating Due to Slip in Heavy-Duty Electric Vehicle Drivetrain Bearings Ocepek, Jakob; Nutakor, Charles; Sopanen, Jussi This work aims to provide further insight into the dynamic behaviour of electric drivetrain ball bearings in heavy machinery, with particular focus on slip occurrence and its impact on localised heating. Bearings are critical components within drivetrains, yet they operate under high loads and speeds that create harsh dynamic conditions, potentially leading to slip and accelerated wear. To investigate this phenomenon, a transient flexible cage bearing model is developed. Furthermore, a detailed local heating model based on the discrete convolution and FFT (DC-FFT) methodology is introduced to evaluate the severity of slip events. Real-world duty cycle data are employed to determine when and how slip occurs during operation. Findings indicate that slip can occur frequently under real operating conditions; however, its severity in terms of localised heating is strongly influenced by local speeds and loads at the slipping contact. Qualitative analysis reveals that, during nominal operation, localised heating remains relatively mild; in contrast, under more severe conditions, peak temperatures can rise to dangerously high levels, posing a risk of damage or even immediate failure.

2026-06-23T00:00:00Z Okolie, Jude A. Khan, Ahsan Khan, Usman http://lutpub.lut.fi:80/handle/10024/173085 2026-10-09T10:00:12Z 2026-10-08T00:00:00Z

Corrosion Mechanisms and Mitigation Strategies during Electrochemical Discharge of Spent Lithium-Ion Batteries Okolie, Jude A.; Khan, Ahsan; Khan, Usman The discharging of spent lithium-ion batteries (LIBs) is a critical pretreatment step in the recycling process, as it minimizes the risk of short circuits, thermal runaway, and fire, thereby ensuring operational safety. However, the battery discharge process frequently induces corrosion of metallic components, which could lead to structural degradation and the potential leakage of toxic electrolytes. In this study, we discuss the underlying corrosion mechanisms that occur during electrochemical discharge of spent LIBs in various electrolyte systems. This discussion consolidates current knowledge on corrosion pathways, the environmental impacts of electrolyte degradation, and mitigation strategies, including inhibitor addition, external isolation, and low-chloride smart solvents. We also outline different metrics and quantitative corrosion characterization approaches. These methods are essential for comparing discharge media and determining whether rapid electrical discharge is achieved at the expense of battery casing, terminal, or current collector degradation. Finally, we outline future directions for developing safer discharge media to enable low-corrosion and environmentally sustainable LIB recycling.

2026-10-08T00:00:00Z Abdelsalam, Ahmed Happonen, Ari Papadakis, Panagiotis Porras, Jari http://lutpub.lut.fi:80/handle/10024/173084 2026-10-09T09:00:11Z 2026-01-01T00:00:00Z

MoRE: Multi-Oracle Reward Evolution for Automated Reward Shaping in Reinforcement Learning Abdelsalam, Ahmed; Happonen, Ari; Papadakis, Panagiotis; Porras, Jari Reward design remains a critical bottleneck in reinforcement learning (RL): handcrafted reward functions require specialized expertise, while automated single-model large language model (LLM) approaches suffer from hallucinations and inconsistent outputs. We introduce MoRE (Multi-Oracle Reward Evolution), a framework that leverages an ensemble of LLMs, each acting as a distinct oracle, to generate and iteratively refine white-box reward functions. Identical prompts are issued to multiple oracles, yielding a diverse candidate pool that is evaluated via GPU-accelerated Proximal Policy Optimization and evolved through in-context multi-model reward reflection. Across 29 tasks in the Isaac Gym and Dexterity benchmarks, MoRE matches or surpasses single-LLM methods, expert-engineered rewards, and sparse baselines, achieving the highest success rates in 24 environments. By reducing human effort, mitigating hallucination risk through model diversity, and scaling reward generation and tuning, MoRE establishes a more reliable and efficient pathway for automated reward design in complex RL control problems.

2026-01-01T00:00:00Z Jaatinen-Värri, Ahti Uusitalo, Antti Tiainen, Jonna Turunen-Saaresti, Teemu http://lutpub.lut.fi:80/handle/10024/173083 2026-10-09T07:30:11Z 2026-10-01T00:00:00Z

A method for predicting the off-design performance of real gas centrifugal compressors utilising preliminary design data Jaatinen-Värri, Ahti; Uusitalo, Antti; Tiainen, Jonna; Turunen-Saaresti, Teemu Being able to predict the centrifugal compressor off-design performance early on in the design process enables designers to match the compressors’ off-design behaviour with the system requirements. For example, they can assess that the compressor is able to produce the required pressure ratios and have a sufficient operating range. In recent years, centrifugal compressors have been gaining increasing interest in applications, such as supercritical CO2 power cycles or high-temperature heat pumps, where the working fluid often exhibits non-ideal thermodynamic behaviour. There are tools in the open literature which can predict the centrifugal compressor operating maps in the preliminary design phase. However, most published tools assume that the working fluid is an ideal gas. This paper presents a method for predicting the off-design performance of centrifugal compressors when the working fluids operate in non-ideal thermodynamic regimes. The approach is based on a previously published method, which is modified. Both methods assume that off-design efficiencies and work-coefficients depend on their design operating point values, the tip Mach number and the flow coefficient at the off-design operating point. The original method assumes the working fluid to be a perfect gas. The presented method predicts the off-design performance of single- or multi-stage centrifugal compressors operating with fluids exhibiting non-ideal thermodynamic behaviour with good accuracy. The mean absolute differences between the measured and modelled pressure ratio and efficiency for air are 0.073 and 0.038, respectively. For supercritical CO2 the differences for the pressure ratio and efficiency were 0.013 and 0.043. When the proposed method is compared with the original one at identical compressor inlet conditions, the difference between the predicted performance increases when the molecular complexity of the working fluid increases. When the number of active degrees of freedom, used to measure the molecular complexity, increase from 5.1 (air) to 24.7 (R1233zd(E)), the difference in the predicted pressure ratio with the two methods increase from 0.03 to 0.79.

2026-10-01T00:00:00Z