Nonlinear Model Based Information Synthesis and Health Detection with Applications to Drive-by-Wire Engines

Purdue University, West Lafayette IN

Investigators

Abstract

PIs: Matthew Franchek and George Chiu, School of Mechanical Engineering, Purdue University Proposal Number: 0097807 Title: Nonlinear Model Based Information Synthesis and Health Detection with Applications to Drive-by-Wire Engines Abstract: This proposed research will create sensor software technology from an Information Technologies (IT) point of view called information synthesis (IS). This knowledge base starts from the premise that sensor and actuation information creates a critical data set which can produce nonlinear models that span the entire information space of a system. The IS technologies will be built by creating an online adaptive nonlinear (NL) modeling technology which is then integrated with analytical functions designed for health detection. The IS technology will be experimentally validated on a drive-by-wire internal combustion (IC) engine research to create fail-safe engines. Research support for the drive-by-wire portion of this work will be sought from industrial sponsors. The scientific impact of this proposed research is the creation of an IS knowledge base applicable to a large class of physical systems. The proposed IS technology complements the data fusion knowledge base by synthesizing information via nonlinear dynamic models instead of I/O tabular maps. Therefore as the system ages, the IS online model adaptation will eliminate the need to create a new I/O mapping. Furthermore, this model based IS approach will allow information extrapolation to a space corresponding to input signals not originally considered. The first contribution will be the creation of a NL dynamic system modeling technology that can be applied to either experimental data or numerical simulations, or it can be extracted from other NL models such as an artificial neural network. To maintain model accuracy over the product life, the second contribution of this research will develop a passive online model adaptation technology. The next contribution will be the development of an analytical design technology for health detection. Specifically, the adapted model coefficients will be analyzed to extract product health. These IS technologies will be validated on a fully electronic drive-by-wire engine. The engine is a Ford V-8 fuel injected engine fitted with an electronic throttle for fueling control and torque control management. The goal here is to create a fail-safe drive-by-wire engine that can withstand sensor failures that would otherwise lead to engine power burst.