Refine
H-BRS Bibliography
- yes (4) (remove)
Departments, institutes and facilities
Document Type
- Article (2)
- Conference Object (1)
- Preprint (1)
Language
- English (4)
Keywords
- Machine learning (4) (remove)
The lattice Boltzmann method (LBM) is an efficient simulation technique for computational fluid mechanics and beyond. It is based on a simple stream-and-collide algorithm on Cartesian grids, which is easily compatible with modern machine learning architectures. While it is becoming increasingly clear that deep learning can provide a decisive stimulus for classical simulation techniques, recent studies have not addressed possible connections between machine learning and LBM. Here, we introduce Lettuce, a PyTorch-based LBM code with a threefold aim. Lettuce enables GPU accelerated calculations with minimal source code, facilitates rapid prototyping of LBM models, and enables integrating LBM simulations with PyTorch's deep learning and automatic differentiation facility. As a proof of concept for combining machine learning with the LBM, a neural collision model is developed, trained on a doubly periodic shear layer and then transferred to a different flow, a decaying turbulence. We also exemplify the added benefit of PyTorch's automatic differentiation framework in flow control and optimization. To this end, the spectrum of a forced isotropic turbulence is maintained without further constraining the velocity field.
The analysis of used engine oils from industrial engines enables the study of engine wear and oil degradation in order to evaluate the necessity of oil changes. As the matrix composition of an engine oil strongly depends on its intended application, meaningful diagnostic oil analyses bear considerable challenges. Owing to the broad spectrum of available oil matrices, we have evaluated the applicability of using an internal standard and/or preceding sample digestion for elemental analysis of used engine oils via inductively coupled plasma optical emission spectroscopy (ICP OES). Elements originating from both wear particles and additives as well as particle size influence could be clearly recognized by their distinct digestion behaviour. While a precise determination of most wear elements can be achieved in oily matrix, the measurement of additives is performed preferably after sample digestion. Considering a dataset of physicochemical parameters and elemental composition for several hundred used engine oils, we have further investigated the feasibility of predicting the identity and overall condition of an unknown combustion engine using the machine learning system XGBoost. A maximum accuracy of 89.6% in predicting the engine type was achieved, a mean error of less than 10% of the observed timeframe in predicting the oil running time and even less than 4% for the total engine running time, based purely on common oil check data. Furthermore, obstacles and possibilities to improve the performance of the machine learning models were analysed and the factors that enabled the prediction were explored with SHapley Additive exPlanation (SHAP). Our results demonstrate that both the identification of an unknown engine as well as a lifetime assessment can be performed for a first estimation of the actual sample without requiring meticulous documentation.