Analytical instruments

An instrument part is judged on what it does not do: not outgas, not deflect, not disturb the measurement.

What the drawing demands

Mass spectrometry, chromatography, electron microscopy and diffraction systems are built from parts that sit inside a vacuum or next to a detector. That puts requirements on the drawing that a general machining quote does not see: surface roughness and cleanliness that decide how much a part outgasses, non-magnetic material choices near a detector, thermal stability across an optical or mechanical path, and assembly tolerances that matter more than any single dimension. Cleanroom handling follows ISO 14644 particle classes, and the calibration a finished instrument ships with rests on ISO/IEC 17025 accredited measurement. Series are small, documentation requirements are not.

What the model learns here

In this sector the model is learning to connect a roughness or cleanliness note to the surface it governs, a material designation to the part that sits near a detector, and a stack of assembly tolerances to the geometry that carries them. Blake proposes these links for review, traceable to the inputs they came from. The craft overlaps with semiconductor equipment, but the customer, the standards and the reason a requirement exists are different, and that difference is exactly what a model has to learn rather than assume.

What the model learns here belongs to the drawing, not to Blake.

Who this matters to

Estimator

Confirms which surface carries a roughness or cleanliness requirement, and whether it forces a finishing pass on top of milling, before pricing the part.

Manufacturing engineer

Plans the handling and finishing a cleanroom-bound part needs, and the fixturing that keeps a thin instrument part from deflecting.

Owner / operations

Carries the risk of a rejected batch: a part that outgasses or shows particles fails at the customer's cleanroom door, not at yours.

Quality / regulatory

Owns the ISO 14644 handling regime and the ISO/IEC 17025 calibration record; Blake only points to where the requirement sits on the drawing.

Where this model can help

Pilot

Linking roughness and cleanliness notes

A surface roughness or cleanliness requirement is proposed as linked to the face it governs, so a vacuum-facing surface keeps its reference, for the estimator to confirm.

Pilot

Tracing material restrictions

A non-magnetic or low-outgassing material note is shown with the same traceability as any other requirement: measured, derived, assumed or unconfirmed.

Direction

Assembly tolerance stacks, direction

How tolerances accumulate across a mounted assembly is something the model is designed to learn to reason about. Not yet confirmed running.

Direction

Instrument parts as a feature family, direction

Vacuum housings, mounts and optical benches share a family of features the model is designed to recognise across drawings. A design principle, not a confirmed, ratified function, and not a claim that Blake is already used in this sector.

Process-specific AI models

One shared representation of the part, its requirements and the factory. Each manufacturing process gets a specialised AI model that learns how geometry, process choices and production resources affect manufacturability, time and cost. Milling first.

Milling

First process model

Features, tooling, setups and cycle time for 3-axis and 3+2 machining.

Turning

Second process model

Turned profiles and grooves on the same feature model.

Wire and sink EDM

Next process family

Extending the world model beyond mechanical cutting into electrical-discharge machining, developed with specialist manufacturing partners.

Sheet metal

On the roadmap

Grinding

On the roadmap

Welding and assembly

On the roadmap