Technical & Laboratory Work

• I investigated meshes up to approximately 3.6 million elements, including adaptive Hessian refinement, while assessing convergence, mesh dependence and computational cost.
• I compared aerodynamic coefficients and angle-of-attack behaviour with analytical and theoretical models, including laminar and turbulent boundary-layer and transition behaviour.
• I compared Spalart–Allmaras and k–ω turbulence models to see how turbulence modelling changed the result and the computational trade-off.

What I took from it:The work developed my ability to treat CFD as an analysis and validation process rather than a single plotted result.

• I studied cantilever beams, plates with stress concentrations and full wing-shell models using 1D, 2D and 3D finite elements.
• I compared linear and quadratic element formulations and conducted mesh-convergence studies to understand the influence of model resolution and idealisation.
• I evaluated wing thickness and material selection against weight, deflection and factor-of-safety constraints. Here, structural optimisation means an engineering trade-off investigation rather than an automated optimisation algorithm.

What I took from it:The progression in model complexity helped me connect idealisation, numerical convergence and practical structural decisions.

• I used state-space, transfer-function and zero-pole-gain representations alongside eigenvalue, time-domain, frequency-domain, Bode and root-locus analysis.
• The reduced model agreed with the full model to approximately 0.96% in natural frequency (0.9171 versus 0.9260 rad/s) and 1.09% in damping ratio (0.5182 versus 0.5239) for the short-period dynamics.
• The SPO approximation could not reproduce the low-frequency phugoid mode. I also investigated sensor lag, Gaussian measurement noise, low-pass filtering, sampling and ADC quantisation rather than treating the controller as an ideal system.
• PID tuning reduced overshoot from approximately 35% to 9%. Stability remained for the tested short-period natural-frequency cases, without implying complete controller robustness.

What I took from it:The useful result was not just a close numerical match: it showed where a reduced aircraft model was useful and where its limitations mattered.

• I measured velocity, pressure, lift and drag across multiple Reynolds numbers and angles of attack, including boundary-layer and pressure-tapped aerofoil work.
• I used numerical integration of pressure distributions to obtain lift coefficients and compared the experimental results with XFOIL, XFLR5 and theoretical models.
• The comparison identified approximately 17% difference in lift-curve slope and 54% difference in glide-ratio prediction. These differences helped identify where simplified drag modelling became less reliable.

What I took from it:This work demonstrates the full evidence loop: experiment, data processing, numerical comparison and critical evaluation of model limitations.