Dynamic models adaptation for a 4 Inj - 2PP common-rail pressure system

The paper is focused on discussing the capability of a NeuroFuzzy, nonlinear predicting structure, with Local Linear Models (LLM) designed for fuel pressure estimation in diesel common-rail (CR) system. Nonlinear dynamic systems like CR are not so easy to model and engineers have often struggled to find the best solution to approximate the input-output dependencies. NeuroFuzzy networks, combined with LLM, are powerful tools for splitting an input space into smaller pieces where the linear approximations are considered satisfactory. Using appropriate numerical models, these architectures can be implemented in a real-time environment with a moderate effort but a new challenge arises: real time adapting of the linear models parameters. The paper illustrates that the LLMs, and hence, the whole dynamic models parameters of the CR´s NeuroFuzzy developed architecture, can be adapted for a wide working space. The practical final results are favorable. The solution may result in lower emissions with favorable economic and environmental implications.