| Summary: | Powertrain applications require high performance controllers yet are in general restricted to low order solutions due to limitations in the software and hardware. There are many well developed robust feedback techniques which can be applied very successfully to automotive systems, however these are generally high order solutions and are therefore not suited to most commercial powertrain control modules (PCM). Typically, powertrain control strategies are implemented through low order look-up tables. There are only a limited number of robust design methods which can be used for fixed, low order controller design and many of these techniques are limited to single-input-single-output (SISO) problems. New technology engines are being developed with additional mechanical systems to increase the performance. As these technologies are developed the interactions between system inputs and outputs are increasingly coupled and therefore it is necessary to consider engines .as multivariable systems. Accordingly, for high performance control it is necessary to move away from single control loops or sequential SISO loop designs favoured until recently. To achieve high performance with the constraints of the PCM this thesis develops a series of 'engineer friendly' controller design tools. Two distinct parameter space (PS) control techniques are detailed, which are particularly suited to low and fixed order control. These methodologies are intended to be suitable for non-control experts by offering insights into the design constraints. The first technique is a novel PS approach to constrained and minimum variance (MV) controller design for both continuous and discrete systems. This technique is successfully applied to the peak pressure position (PPP) control problem using spark advance (SA) as the input. The second technique developed is a multivariable 'Hoc parameter space technique for designing fixed, low order controllers. This technique uses only frequency response information in the design scheme and is therefore equally suited to both continuous and discrete systems. Controllers are developed by a series of parameter plane iterations, which can be used for re-tuning controllers from alternative design methods. Equally, the technique can be used for the direct design of controllers, starting with no initial transfer function gains. The method is successfully demonstrated as a direct design technique for a single sensitivity problem with time response criteria. A further example demonstrates the technique for weighted sensitivity reduction using the direct design approach and also for re-tuning a reduced order controller obtained by an algebraic Riccati design. To facilitate the application of the multivariable PS technique, a novel approach to designing a feedforward fuelling controller for a port fuel injection (PFI) gasoline engine is detailed. The feedforward controller is based on inverse nonlinear auto-regressive moving average (NARMA) ordina.ry least squares (OLS) identification. The approach is applied successfully to the idle speed region of an engine and has the advantage on linearising the plant dynamics. The system with compensator is subsequently re-identified and coupled with a robust multivariable PS controller for control of the idle speed and air-to-fuel ratio (AFR).
|
|---|
