Development and analysis of adaptive neural network control for a Cybernetic intelligent ‘iGDI’ engine

As combustion can widely vary between engine cycles if left uncontrolled, strict and robust control methodologies required to meet optimum performance at different operating conditions. Although the direct injection stratified charge engine development attempts started since 1920, lack of advanced instrumentation and robust real-time control techniques restricted its developments to common usages. In this research, novel intelligent control solutions developed and simulated on a Gasoline Direct Injection (GDI) engine. A research four cylinder 2.0 L GDI engine modeled with specific mechanical hardware along with adaptive control software that is frequently called as the conceptual Cybernetic intelligent GDI or ‘iGDI’ engine The engine modeled with Free Valve Actuation (FVA) hardware and precision fuel injectors connected directly to the engine cylinders that found assistive for control flexibility by technical assessments. Then a mechatronic neural network control approach proposed with adaptive control techniques. The engine and the controllers modeled and simulated with GT-SUITE and SIMULINK coupled simulation for control performance validation. Adaptive and predictive neural control architectures developed for multiple distinct GDI combustion modes. High volumetric efficiency (~99%) obtained overcoming pumping loss with throttle-less drive-by-wire operation. The neurocontrollers trained for time varying plant dynamics with Nonlinear Autoregressive with eXogenous Input (NARX) neural network. AFR set-point tracking achieved on ‘ECO’ mode with NARMA-L2 controller for minimum BSFC and NOx. On ‘POWER’ mode operation, maximum brake torque (MBT) obtained with NN predictive controller. Intake boosting provided by single stage turbocharging with intelligent controller methods developed and discussed. NN based controller algorithms developed, trained and simulated for dynamic control on cylinder firing events named as intelligent Dynamic Skip firing (iDSF) operation. Performance upgrades with the new hardware and software solutions discussed and shown graphically. It is clearly resulted that computational intelligence could effectively handle highly nonlinear dynamic real-time MIMO engine control problem with advantages of online optimization and adaptation feature.