A Proposed Methodology for Biodiesel Blend Impact Evaluation on Engine Emissions Using Gaussian Process Modelling

The growing concern with environmental impact is a major drive in the search of alternatives to conventional fossil fuels. In this context, the blending of some quantity of biodiesel in diesel as a mean of reducing CO2 emission levels is being increasingly adopted. Brazil, for example, has determined a steady increase from 2% (2008) up to 12% (2020) in blend rate with no signs of stop or reversion on the tendency. Nevertheless, the expected impact of this fuel evolution on engine pollutant emissions raises questions on how an ageing Diesel fleet will react and how new calibrations may be efficiently developed.A promising approach to tackle this challenge in a cost-effective way is the adoption of Gaussian process regression (GPR) in the context of model-based calibration, using virtual, use-case specific, driving cycles. The aim of this paper is to evaluate this methodology on São Paulo context by selecting a four cylinder, Euro V diesel engine, representative of the city bus fleet, and combining it with a São Paulo’s urban bus driving cycle and 10% biodiesel blend rate.Firstly, a space-filling test plan was constructed aiming to explore parameter combinations on the neighborhood of the original calibration. The proposed test cases were measured on an engine dynamometer and then fed to ETAS ASCMO (Advanced Simulation for Calibration, Modeling and optimization). The tool generated a virtual engine model based on Gaussian process regression with which the calibration was optimized for each different cycle