A nine-year descent from toolpath to process physics

Problem

On a high-volume line, a defect mode that is understood only statistically will recur. Escapes and yield loss get managed by reaction rather than removed at the cause.

What I did

Investigate defect modes down to their physical mechanism, then convert each confirmed cause into a process control rather than an inspection step.

Why this way

Inspection detects; control prevents. At gigafactory volume the economics of detection are unfavourable — every part inspected is a part already made, so the leverage sits upstream in the process window.

Validation

Each change verified against the line's own SPC record over a sustained production window, then locked into the control plan and re-checked at process audit.

Problem

Battery cell leak testing was performed at a manual station. Throughput was operator-limited, and escape risk tracked operator attention across a full shift rather than the capability of the test itself.

What I did

Designed an automated leak-detection cell — pneumatic sealing fixture, sensing, and a robot for part transfer — and specified the control sequence for safe, repeatable operation.

Why this way

A fixture-plus-robot cell was chosen over a fully custom indexing machine because cell format changeover was expected. A robot absorbs geometry changes in software; a bespoke indexer absorbs them in retooling.

Validation

Verified against calibrated reference leak specimens at machine buy-off, then re-verified on-line during production ramp.

Case study from this role

• Battery Leakage Detection Machine

Problem

Magnesium alloys offer the lowest structural density of the engineering metals but limited stiffness and energy absorption. Adding reinforcement helps only if it can be distributed uniformly without destroying castability.

What I did

Developed an infiltrated stir casting route for a magnesium hybrid metal composite, then characterised the resulting microstructure and measured its energy-absorption behaviour.

Why this way

Infiltration combined with stirring was chosen to address the specific failure mode of conventional stir casting — reinforcement clustering and settling — while remaining a route that a foundry could realistically reproduce.

Validation

Peer review. Process route and results published in the Journal of Applied Science and Engineering, 29(2), 2026.

Case study from this role

• Magnesium Hybrid Metal Composite

Problem

Agro-processing plants run continuously in abrasive, dust-laden conditions. Machinery that is designed for peak throughput but not for maintenance access loses more capacity to downtime than it ever gained in rate.

What I did

Designed a family of process machines — rotary dryers, conveyors, palletizers, and dust collectors — sizing each for continuous duty and detailing them for serviceability.

Why this way

Specified conservative drive and structural margins and prioritised access for wear-part replacement, because in this duty cycle availability dominates rated capacity.

Validation

Commissioned on site and handed over against the client's throughput and continuous-operation requirements.

Case study from this role

• Agricultural Machinery Systems

Problem

Mold design decisions are committed before the first shot and are expensive to reverse. Gate position, cooling layout, and ejection have to be resolved against part quality and cycle time simultaneously.

What I did

Designed mold tooling with gating, cooling, and ejection resolved against the part's expected shrinkage and the customer's cycle-time target.

Why this way

Prioritised cooling symmetry over the simplest machining route, because uneven cooling shows up as warpage in every part produced for the life of the tool, while a harder-to-machine core is a one-time cost.

Validation

First-article inspection against drawing tolerances, then cycle-time confirmation in production trials.

Problem

On precision tooling, the toolpath decides both the surface finish and whether the part survives. Aggressive strategies save hours and lose parts.

What I did

Programmed multi-axis toolpaths and tuned strategy, stepover, and engagement for each feature class rather than applying one global setting.

Why this way

Matched cutting strategy to feature geometry and material because a single conservative setting wastes machine hours on open features, while a single aggressive one damages thin walls and deep pockets.

Validation

Dimensional and surface inspection after machining, with tooling proven in subsequent moulding trials.