Neural Network Based Model Predictive Controller Design of a Spark-Ignition Internal Combustion Engine CEAS Western Michigan University

Format:

Book

Conference/Event

Author/Creator:

Abotabik, Muataz, author.

Contributor:

Meyer, Rick

Proctor, Christopher

Conference Name:

WCX SAE World Congress Experience (2020-04-21 : Detroit, Michigan, United States)

Language:

English

Physical Description:

1 online resource cm

Place of Publication:

Warrendale, PA SAE International 2020

Summary:

Internal combustion engine (ICE) control techniques have been developed with only the consideration of the first law of thermodynamics, e.g. improving thermal efficiency, tracking specific load requirements, et cetera The first law of thermodynamics doesn't account for the losses in work potential that are caused due to the in-cylinder high temperature thermodynamic processes irreversibilities. For instance, up to 25% of fuel exergy or fuel availability may be lost to irreversibilities during the combustion process. The second law of thermodynamics states that not all energy in an energy source is available to do work; its application evaluates the maximum available energy in that source after accounting for the losses caused by the irreversibilities. Thus, including the exergy in an optimal engine control algorithm may lead to improved ICE thermal efficiencies. In this work, a model predictive controller (MPC) is developed based on the first and the second law of thermodynamics to control a detailed eight-cylinder ICE developed in GT-Power. To make the controller practically applicable for eventual hardware in the loop (HiL) investigations, the detailed GT-Power model is approximated with a single layer feedforward neural network (SLFN) that was trained on engine maps developed from a design of experiments. The MPC optimization problem is solved by two approaches: sequential quadratic programing (SQP) and continuation/forward difference generalized method of residuals (C/FDGMRES) for the purpose of comparing solution time and performance. Results for MPC that incorporates the second law shows fuel saving of 6.8% and 3.2% for SQP and C/FDGMRES, respectively, compared to first law of thermodynamics-based results. C/FDGMRES solved the control problem on average four times faster than SQP

Notes:

Vendor supplied data

Publisher Number:

2020-01-0250

Access Restriction:

Restricted for use by site license