Projects

Achievement: Engineered a robust 9 stage configuration using the free vortex law to maintain radial equilibrium while keeping a constant mean aerodynamic diameter of 0.64 m. I implemented a non uniform enthalpy drop distribution, deliberately offloading the critical first stage and increasing work progressively through the machine. To ensure a smooth transition of flow angles, I formulated and numerically solved an 8th degree polynomial equation for the stator exit angles, identifying an optimal decay coefficient (k) of 0.843.

Insights: Through iterative solidity optimisation, I constrained the Lieblein Diffusion Factor below the 0.45 safety threshold for every radial section. The analysis required careful management of transonic flow conditions at the first stage tip, where the Mach number reached 1.26. I balanced this by fine-tuning blade camber and momentum thickness ratios. The final design successfully synchronised 17 rotor and 25 stator blades for the initial stage, providing a high stall margin and a continuous, shock-free velocity triangle progression along the entire compressor axis.

• Theory: Radial equilibrium via free vortex law and 1D/2D meanline parameterisation
• Modelling: Variable multi-stage enthalpy distribution and 8th degree polynomial numerical solver
• Validation: Lieblein diffusion limits, Mach number mapping, and aerodynamic efficiency characterisation

Achievement: Developed a numerical solver using the Method of Characteristics to compute the flow field downstream of a 0.1 m radius throat. I built a framework to solve the compatibility equations for both positive and negative characteristics, iteratively determining the intersection points and local Mach numbers. The solver was designed to target an isentropic area ratio of 4.235, which corresponds to the theoretical expansion required for a Mach 3 exit condition in air with a specific heat ratio of 1.4.

Insights: Conducted a rigorous convergence study to optimise the discretisation of the characteristic net. By comparing 10, 20, and 79 characteristic lines, I demonstrated that 79 lines represent the optimal threshold where the area error converges below 0.01 percent. The analysis confirmed that increasing the discretisation density not only improves the fidelity of the exit radius (targeting 0.206 m) but also significantly reduces the total axial length of the divergent section, providing a more compact and efficient propulsion component.

• Theory: Prandtl Meyer expansion and compatibility equations in supersonic flows
• Algorithms: Custom MOC solver with iterative intersection point refinement
• Validation: Area ratio convergence and Mach 3 exit profile uniformity checks

Achievement: Implemented a hierarchical design process. I began with a 1D meanline approach to estimate hub and casing radii, followed by an axisymmetric throughflow calculation to resolve the radial equilibrium. I then used a quasi-orthogonal 3D (Q3D) iterative analysis to optimise individual blade profiles before performing a full 3D Navier-Stokes analysis on the final stacked blades using JD75.

Insights: By iterating on stage work fractions and inlet swirl angles, I successfully eliminated zones of negative reaction that were identified in the early throughflow stages. The final 3D validation demonstrated that while Q3D methods effectively capture profile losses, the high-fidelity 3D RANS was required to accurately quantify the secondary flow losses and endwall vortices that dictate the true efficiency of high-pressure turbine stages.

• Methodology: Meanline, Axisymmetric Throughflow, and Q3D profile optimisation
• CFD: 3D RANS analysis of stacked blades using JD75 (Denton code)
• Constraints: Positive stage reaction management and Mach number limit enforcement

Achievement: Employed a multi-fidelity design process. Analysed boundary layers via a 2D viscid-inviscid solver (MISES) and performed 3D CFD (MULTALL) to optimise blade lean.

Insights: Trained a neural network surrogate model on the CFD database, allowing for rapid, low-cost prediction of the critical trade-off between loss coefficients and pressure rise.

• CFD: 3D RANS (MULTALL), 2D coupled viscid-inviscid solver (MISES)
• Machine Learning: Neural network surrogate modelling
• Optimisation: Multi-objective trade-off analysis (loss vs. pressure rise)

Achievement: Conducted 0D thermodynamic cycle analysis to select optimal pressure ratios and bypass ratios for the specific flight envelope.

Insights: Performed meticulous meanline sizing and component matching of the compressor, combustor, and turbine to meet thrust and specific fuel consumption targets.

• Analysis: 0D thermodynamic cycle assessment
• Design: Meanline component sizing and aerodynamic matching
• System: Full propulsion system integration

Achievement: Conducted a complete aerodynamic synthesis. I used Balje and Jaumotte statistical diagrams to select the optimal specific speed and dimensionless pressure and flow coefficients. The 11-blade impeller geometry was mathematically generated using the Kaplan error triangle method to transition from planar coordinates into cylindrical space.

Insights: Rather than assuming ideal Stepanoff conditions, I wrote a MATLAB script to plot the actual constant specific speed curve and find the true intersection with the operational line. This allowed me to correctly adjust the theoretical efficiency drop caused by the machine's smaller scale, ensuring the final volute and divergent diffuser sizing was grounded in realistic flow physics.

• Theory: Stepanoff diagrams, Balje and Jaumotte specific speed matching
• Design: Kaplan error triangle method for blade profiling, volute section sizing
• Algorithms: MATLAB non-linear curve intersection to find operating points

Achievement: Engineered a 9-stage compressor configuration by solving an 8th-degree equation for deviation angle decay. Built a numerical solver for the nozzle using the Method of Characteristics.

Insights: Mapped the Prandtl-Meyer expansion waves for the nozzle and proved that computing exactly 79 characteristic lines provided the strict mathematical threshold to minimise area error for a uniform Mach 3 exit flow.

• Compressor: Meanline analysis, velocity triangles, diffusion factor validation
• Gas dynamics: Supersonic expansion, shock-free design
• Algorithms: Method of characteristics iteration

Achievement: Engineered an explicit steady-state solver using a cell-centred finite volume discretisation. I implemented a multi-stage Runge-Kutta time integration scheme for enhanced stability and a deferred correction method to achieve second-order spatial accuracy. To maximise computational efficiency, I integrated spatially variable time-stepping based on the local Courant-Friedrichs-Lewy (CFL) condition.

Insights: The primary optimisation involved the calibration of the Jameson-Schmidt-Turkel (JST) artificial viscosity. By fine-tuning the smoothing factors, I successfully balanced shock-capturing capability with global entropy conservation. The solver was validated on a complex supersonic bend case, reaching a mean Mach number of 3.03 at the outlet with clean residual abatement and convergence achieved within the first 100 iterations.

• Algorithms: 4-stage Runge-Kutta integration, cell-centred finite volume method
• Stability: JST artificial viscosity and local CFL condition management
• Performance: Spatially variable time-stepping and high-order deferred correction

Achievement: Extracted accurate CAD geometry and used CFD to quantify the air multiplier effect and jet entrainment ratios.

Insights: Mapped how internal flow path geometry and the Coanda effect directly influence exit velocity profiles and the resulting acoustic signature.

• Modelling: Reverse engineering and CAD geometry extraction
• Fluid Mechanics: Jet entrainment physics, Coanda effect analysis
• Acoustics: Acoustic signature correlation

Achievement: Mapped the full operational envelope of a Solar & Palau TD-500/160 elicocentrifugal fan at 1850 and 2500 RPM. I managed the instrumentation suite, including Pitot-static probes and Betz micro-manometers, and used a calibrated nozzle to determine precise volumetric flow rates through the circuit.

Insights: By computing the dimensionless flow and pressure coefficients, I proved that Similarity Theory holds strictly only at high flow rates where Reynolds number effects are negligible. To model multi-fan systems, I developed an iterative solver in MATLAB to resolve a 5th-degree polynomial characteristic equation, allowing for the precise prediction of flow distribution in complex parallel configurations.

• Testing: Performance characterisation via Betz micro-manometers and DPI 610 sensors
• Analysis: Dimensionless similarity mapping and Reynolds effect quantification
• Modelling: 5th-degree polynomial iteration for parallel system impedance matching

Achievement: Built an hour-by-hour operational model. Optimised the PV array layout and tilt to minimise shading using UNI 10349 standards. For the CSP variant, I sized the parabolic trough solar field and designed a 6-hour thermal energy storage tank (holding over 1.3 million kg of oil) to guarantee autonomy for the Organic Rankine Cycle (ORC) power block.

Insights: By accurately modelling Incident Angle Modifiers (IAM) through custom polynomials and setting a strict 40 MWh minimum activation threshold for the ORC, I proved that while the CSP plant offered dispatchability, it yielded a highly negative Net Present Value. The analysis demonstrated that the PV architecture offered a superior break-even trajectory due to significantly lower CapEx.

• Modelling: Solar irradiance calculation, IAM polynomials, and shading analysis
• Design: PV layout, parabolic trough sizing, 6-hour thermal storage integration
• Economics: NPV calculation, 25-year incentive tracking, CapEx cost stacking

Achievement: Built a full-stack physical and financial model. Processed 8760 hourly data points to compute the specific wind shear exponent and extrapolate hub-height velocity. I corrected for site-specific air density (150m a.s.l.) and modelled turbine performance by interpolating manufacturer data via a custom 4th-degree polynomial.

Insights: The multi-variable analysis revealed a counter-intuitive financial strategy. By deliberately sizing the farm down to three 20 kW turbines (remaining under a 60 kW regulatory threshold), the project secured a much higher feed-in tariff, achieving a superior Internal Rate of Return and a 10-year payback period compared to larger multi-megawatt setups. I also mathematically verified that the array required a 2.25 km setback distance to satisfy spherical divergence noise limits.

• Resource assessment: 8760-hour wind shear power-law extrapolation
• Performance: 4th-degree polynomial power curve integration
• Economics: NPV, IRR, and regulatory feed-in tariff optimisation
• Environmental: Acoustic propagation and noise constraint mapping

Achievement: Derived the h-Q flow rating curve and performed a complete hydraulic analysis. I calculated penstock diameters by optimising the trade-off between Darcy-Weisbach friction losses and capital pipe costs over a 120 m head. The study compared two different turbomachinery architectures: a high-head Pelton wheel and a reaction Francis turbine.

Insights: By calculating the Net Positive Suction Head (NPSH) to prevent cavitation and matching the specific speeds, the techno-economic analysis identified a 1.71 MW Francis turbine as the optimal solution. The integrated financial model proved this configuration maximised annual energy production, yielding a payback period of just 7 years with a profitability index of 1.24.

• Hydraulics: Penstock sizing, Darcy-Weisbach head loss, cavitation analysis
• Machinery: Turbomachinery selection (Pelton vs Francis), specific speed matching
• Economics: Net present value, internal rate of return, 7-year payback projection

Achievement: Conducted theoretical characterisation using Classical Lamination Theory (CLT). Validated the stiffness matrices for layups like [0, 30, -30]s against experimental data.

Insights: Quantified the coupling between extension and bending in non-symmetric laminates. Observed experimental stiffness values (~21 GPa) matching orthotropic predictions, defining optimal layup sequences to prevent delamination.

• Theory: Composite mechanics, off-axis stress analysis
• Modelling: Finite element analysis of NACA profiles
• Application: Structural resilience against peak hydrodynamic bending moments

Achievement: Executed a complete aerospace-grade manufacturing cycle using hand lay-up, vacuum bagging, and autoclave consolidation. I then conducted destructive tensile and flexural testing campaigns on an Instron 5585 universal testing machine, using precision MTS extensometers to capture multi-axial strain data.

Insights: By testing specific stacking sequences (such as using a [+45/-45]2s laminate to isolate the in-plane shear modulus), I successfully extracted the fundamental stiffness parameters. During the flexural testing phase, I performed variable-span 3-point bending tests (from 160 mm down to 30 mm) to explicitly quantify the scale at which transverse shear deformation begins to artificially reduce the apparent flexural stiffness, proving the necessity of high span-to-thickness ratios for accurate modulus characterisation.

• Manufacturing: Autoclave temperature and pressure cure cycles, vacuum bagging
• Testing: ASTM standard tensile and flexural destructive characterisation
• Analysis: Orthotropic elasticity extraction, transverse shear influence mapping

Achievement: Engineered and computationally optimised a 15-layer asymmetric carbon-epoxy laminate stack to precisely balance longitudinal stiffness with shear strength. I executed the physical manufacturing using pre-preg tube rolling around a Teflon mandrel, applying 2.52 bar of radial consolidation pressure via thermodynamic contraction using Hi-Shrink Tape at 120 degrees Celsius.

Insights: Developed a 7-step progressive damage model in MATLAB to simulate ply-by-ply matrix degradation under increasing torque. By recalculating the reduced stiffness matrices after each localised failure, the model successfully predicted the exact sequence of ply failures and the non-linear angular deformation trajectory prior to ultimate catastrophic breakage.

• Mechanics: Torsional buckling, flexural critical speed, Tsai-Wu interaction criteria
• Simulation: 7-step progressive damage modelling, localised matrix cracking degradation
• Manufacturing: Pre-preg tube rolling, thermal shrink-tape consolidation

Achievement: Developed a MATLAB codebase to automate Classical Lamination Theory (CLT). The program calculated the extensional, coupling, and bending stiffness matrices (the [A], [B], and [D] matrices) from constituent fibre and matrix properties across various stacking sequences. I then benchmarked these theoretical predictions against experimental data from tensile and bending tests.

Insights: The study explicitly quantified the coupling effects inherent in non-symmetric laminates, such as extension-bending and bending-twisting interactions. By integrating the Tsai-Hill and maximum stress failure criteria, the model successfully predicted the first-ply failure loads of coupled laminates with a deviation of less than 10% from the experimental rupture data.

• Theory: Classical Lamination Theory, stiffness matrix derivation
• Failure criteria: Tsai-Hill envelope, maximum stress criterion
• Mechanics: Anisotropy, extension-bending coupling quantification

Achievement: Instrumented a cylindrical aluminum vessel with foil strain gauges in a quarter-bridge Wheatstone configuration. I used micrometric tools to characterise the vessel's geometric tolerances and executed a controlled depressurisation cycle to measure the resulting elastic recovery strain.

Insights: I developed an uncertainty model using Student's t-distribution to account for gauge factors and geometric variance. By applying Mariotte's formula to the measured circumferential strain, I calculated an internal pressure of 0.665 bar. The high precision of the result validated the measurement chain and provided a baseline for the design of industrial safety monitoring systems for pressure vessels.

• Instrumentation: Foil strain gauges and Vishay signal conditioning
• Theory: Mariotte's formula for thin-walled pressure vessel mechanics
• Statistics: Uncertainty propagation via Student's t-distribution

Achievement: Conducted monoaxial tensile tests on standard and holed specimens using an MTS testing suite equipped with 647 hydraulic wedge grips and a 634 1IF extensometer. I performed a rigorous statistical characterisation of specimen geometry through ten discrete measurements per dimension, applying a Student t distribution at a 68.27 percent confidence level to propagate geometric uncertainties into the final calculations.

Insights: To ensure high fidelity results, I implemented a linearity assessment using a normalised Chi squared algorithm in MATLAB. This allowed me to select the optimal 800 point interpolation interval within the elastic regime, successfully converging on a Young Modulus of 70216.95 MPa. The study proved that even with high precision MTS instrumentation, geometric tolerance variance remains the dominant source of experimental uncertainty.

• Testing: ASTM D638 tensile characterisation via MTS hydraulic systems
• Statistics: Uncertainty propagation via Student t distribution and Chi squared checks
• Metrology: High precision extensometry and specimen geometric tolerance mapping

Achievement: Conducted a cantilever bending test on a rectangular specimen instrumented with a six element strain gauge rosette. I used a P 3500 Strain Indicator to acquire multi axial data across half bridge and full bridge configurations. To ensure consistency, I applied a geometric correction factor to align the longitudinal and transverse strain measurements at the same effective distance from the load point.

Insights: I implemented a RANSAC (RANdom Sample And Consensus) algorithm in MATLAB to clean the experimental dataset. The algorithm successfully identified and rejected outliers in the strain curves caused by small mass differences, such as the 1.20g and 3.66g increments, that fell below the effective resolution of the sensor. This statistical refinement converged on a precise Poisson ratio of 0.2902, validating the measurement chain against theoretical expectations for aluminum.

• Instrumentation: P 3500 Strain Indicator and multi axial resistance strain gauges
• Algorithms: RANSAC iterative fitting and outlier rejection for noise reduction
• Analysis: Navier bending formula application and uncertainty propagation

Achievement: Engineered a measurement chain using foil resistance strain gauges in a quarter bridge Wheatstone configuration linked to a P 3500 Strain Indicator. The protocol required critical surface preparation including thermal conditioning to prevent condensation, controlled abrasion, and precise axial alignment to ensure the capture of pure circumferential strain data.

Insights: By applying Mariotte's formula for inverse stress calculation, I translated a measured micro strain of 252 into a final internal pressure of 0.665 bar. The analysis was supported by a rigorous statistical treatment using a Student t distribution for specimen characterisation, achieving a final uncertainty of 1.6 percent and validating the reliability of strain based monitoring for non invasive structural assessment.

• Sensing: Foil sensor installation with lead soldering and bridge balancing
• Physics: Pressure vessel mechanics and membrane stress derivation
• Statistics: Micrometric tolerance analysis and measurement error propagation

Achievement: Conducted a series of dynamic tests across three different clamping configurations, including high rigidity setups with additional metal plates, to measure natural frequency and logarithmic decrement. I used a longitudinal strain gauge in a quarter bridge configuration connected to a P 3500 Strain Indicator. To ensure data fidelity, I implemented a custom zero offset algorithm in MATLAB to correct for signal drift and isolate the true peak to peak amplitude of the harmonic decay.

Insights: By solving the characteristic equation of damped motion and applying the natural frequency formula for inflected beams, I determined a dynamic Young Modulus of approximately 74 GPa. The study revealed that while the elastic modulus remains consistent, the damping coefficient n is highly sensitive to the clamping stiffness, proving that increased constraint rigidity directly correlates with higher observable damping in the experimental signal.

• Testing: Damped free vibration analysis with high frequency strain acquisition
• Modelling: Second order differential equation for damped harmonic oscillators
• Analysis: Sensitivity study on clamping rigidity and boundary condition effects

Achievement: Engineered a robust 3D metrology pipeline using digital projection and high-speed imaging. I implemented and compared two encoding algorithms in MATLAB: Binary Sequential Coding (Gray Code) for absolute phase identification and Continuous Varying Color Code (RGB phase shifting) for superior spatial resolution.

Insights: Developed custom triangulation algorithms to transform pixel-space phase maps into a global coordinate point cloud. The investigation revealed that while the RGB phase shifting method increased data density by an order of magnitude, it required advanced chromatic aberration calibration to maintain sub-millimeter geometric accuracy across complex surface gradients.

• Techniques: Gray Code and RGB phase-shifting codification strategies
• Algorithms: Optical triangulation and mathematical phase unwrapping
• Analysis: Chromatic aberration correction and point-cloud density optimisation