Mechanical Engineering - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemOptimisation of closed-loop aerodynamic systems( 2018)The design of a missile system is a multi-disciplinary engineering activity that involves structural, aerodynamics, rocket propulsion, guidance, electronic, and closed-loop control engineering to name a few. In modern engineering practice, a systems engineering approach is utilised to manage the design of a missile, but this does not necessarily guarantee that the final design is optimal. The process may also be inefficient, requiring many iterations of design, prototyping and testing in order to achieve the required specifications. In this thesis, multi-disciplinary optimisation frameworks are developed that target the aerodynamics and closed-loop control system of a supersonic tail-fin controlled missile. The aerodynamics and control system are highly coupled systems, but it is rare to see these subsystems optimised together in the literature. This is due in part to the computational requirements of the aerodynamic simulations and in part due to many control system design techniques that tend to treat the missile dynamics as immutable. A model representing a supersonic tail-fin controlled missile is developed. The model utilises computational fluid dynamics (CFD) simulations in order to capture the aerodynamic behaviour and a state-space model for the dynamics of the missile. Control algorithms are utilised to perform the autopilot function of the missile. This model serves as a basis on which the aerodynamic shape and controller gains can be optimised. Aerodynamic shape optimisation problems typically have large computational demands thus making them impractical to be used with global optimisation algorithms. The first optimisation framework developed is based on sample based global extremum seeking. It is shown that under certain conditions, the convergence behaviour of CFD simulations can be viewed as plant dynamics and thus extremum seeking techniques can be applied to find the optimal aerodynamic shape. The results are a step toward obtaining globally optimal solutions within comparative computation times of gradient-based optimisers. While useful for shape optimisation, the previous result would still struggle with combined aerodynamic shape and control optimisation problems. The next framework proposed is an adjoint-based gradient optimisation framework. The adjoint method has previously been utilised for static shape optimisation problems, but the result presented here is an extension for dynamic and controlled missile problems. The result shows that with appropriate time-scale separation between the actuator and slow states of the missile, the gradient of the cost function can be found with just two times the computational requirements of mapping the aerodynamic characteristics of the missile. This computational requirement is independent of the number of shape design variables and thus shows its practicability. An example of a missile tail-fin profile and autopilot gain optimisation problem is presented. There exists limitations of the adjoint based framework which prevent its use for certain missile geometries. Consequently, an implicit filtering framework is utilised in combination with the adjoint framework to cater for general missile geometries while still maintaining competitive computational speeds. This framework shows that general missile problems can be optimised without restriction. A number of optimisation examples involving a missile tail-fin profile and platform, missile nose cone and autopilot gains are presented. Lastly, goal-oriented mesh adaptation which has often been utilised in the CFD community to refine their computational meshes is utilised in non-linear model predictive control (NMPC). Goal-oriented mesh adaptation is a result derived from the adjoint method. The control algorithm that is developed is computationally faster than the standard NMPC and therefore can be utilised in so-called "fast" systems.