|POLLOK ALEXANDER JOSEF||Cycle: XXX |
Section: Systems and Control
Tutor: BASCETTA LUCA Major Research topic
:Modelling and Control of Aircraft Environmental Control Systems
Advisor: CASELLA FRANCESCOAbstract:
Unseen by the passengers, aircraft environmental control systems are complex thermodynamic systems, requiring a large quantity of power. The tasks of designing, modelling, optimising and controlling all these systems leave many degrees of freedom to the respective expert, and typically require many design loops to arrive at satisfactory results. This thesis contributes to multiple aspects of this process.
In aircraft environmental control systems (ECS), limit cycle oscillations (LCO) can occur. Those are problematic since the life expectancy of the ECS is affected. Using an equation-based object-oriented modelling language (EOOML), a complete, detailed and dynamic simulation model of an ECS is developed for the first time. This model includes the engine bleed air system (EBAS), the air conditioning pack, the cabin and ducting dynamics as well as the recirculation system. Using simulations, it is shown that LCO occurring in ECS cannot be explained by Helmholtz resonance effects.
To further investigate the cause of the LCO, electropneumatic valves - as used in EBAS - are modelled in more detail, using the Lund-Grenoble friction model. Using this model, LCO in aircraft ECS are predicted for the first time. Several control strategies are devised, implemented and evaluated against this model. A strategy based on a combination of feed-forward control, feed-back control and online tuning of the integral action outperforms all other candidates. A 46% reduction of the developed objective function is achieved when compared to the state of the art.
Current architectures for aircraft cabin climatisation only allow for a small number of temperature zones. Differences in heat load, generated for instance by nonconforming seating class layouts, cannot be compensated by the control system. A new architecture is presented that allows for an infinite number of temperature control zones - at the cost of a more involved control system. Suitable control strategies, as well as failure management strategies are demonstrated.
For optimisation studies in the context of simulation models, a class of controllers is found, based on boundary layer sliding mode control. These controllers do not require any tuning effort and show good performance for many systems during simulations. The high sensitivity to noise is unproblematic, as the system is purely virtual at this stage. These features make them suitable for modelling experts at development stages where the architecture design is not yet finalised.
On the most basic level, usability aspects of EOOML are explored. It is found that the use of inheritance can severely retard the understanding of simulation models. Results also suggest graphical representations to be superior to block diagrams with equation-based and algorithm-based representations taking the middle spot.