X-Ray Scanner for Lumafield

An industrial X-ray CT scanner is a shielded machine, and confirming that the shielding performs as designed means measuring the radiation field around it rather than trusting the drawing. That verification is routine, repeated, and consequential — it is the measurement that lets a team state plainly how a system behaves in the space around it.

The value of automating it is not speed alone. A measurement that a machine performs the same way every time can be compared across units, across builds, and across time. That turns a survey into a dataset.

Today an operator traces a Thermo Fisher RadEye G20 survey meter slowly across the faces of the system by hand. It works, and experienced operators are good at it — but the path, the standoff distance, the dwell time and the coverage all depend on the person holding the instrument.

Two consequences follow. The scan takes a meaningful amount of an engineer's time, and results carry operator-to-operator variation that is difficult to separate from the thing being measured. Both are problems a motion system can address directly.

My first job as technical lead was to turn a measurement task into engineering requirements. What surfaces must be reached, at what standoff, in what orientation, and how precisely does the instrument need to return to a point for two scans to be comparable?

That produced the envelope: a working volume of roughly 7 ft × 7 ft × 4 ft around the scanner, reach to multiple faces without repositioning the frame, and a target positioning repeatability of 1 mm. Repeatability — returning to the same point — matters more here than absolute accuracy, because the comparison of interest is between scans.

The system uses five motorized axes: gantry-scale linear travel to cover the volume, plus the additional degrees of freedom needed to hold the detector square to a surface as the path wraps around corners and across faces of differing orientation.

Choosing where to spend axes was the central architectural tradeoff. Every axis adds mass, cabling, cost and a stack-up of compliance that the end effector inherits. Axes that only exist to serve a small region of the path were candidates for removal in favor of fixturing or a re-run at a second orientation.

At two metres of span, structure stops being a detail. A frame that is adequately strong can still be inadequately stiff: deflection under the moving mass, and sag that changes with position, both land directly on the measurement point as error.

So the stiffness budget is allocated before the aesthetics of the frame are settled. Span, section, bearing spacing and the location of the drive relative to the load are the levers. Alignment is treated the same way — the frame is designed to be squared and re-squared as a deliberate procedure, not assumed to arrive true from assembly.

The end effector holds a Thermo Fisher RadEye G20 — an instrument the team does not control and must not modify. It has to be held repeatably, at a known offset from the tool point, and be removable for calibration by someone wearing gloves.

Cable management is a first-order mechanical problem on a five-axis machine, not a finishing task. Cable carriers add drag and mass exactly where the system is least stiff, and a cable that snags is a failed scan and a damaged instrument. Routing is designed with the motion envelope, not after it.

Coverage is the point of the machine. Paths are planned so the instrument sweeps each face with consistent pitch, standoff and dwell — the parameters that make one scan comparable to the next.

The mechanical design and the path plan constrain each other. A path that demands high acceleration in the least stiff part of the envelope is a path that will not repeat, so path design is treated as part of the mechanical design rather than a downstream software concern.

The plan is to prove the mechanics before trusting the measurement. Repeatability is characterized as a mechanical property first — commanding the machine to return to a set of points and measuring where the tool point actually lands, independent of any radiation reading.

Subsystems are prototyped ahead of the full frame: the end-effector interface, a single axis at representative span, and the cable routing through its full travel. The intent is to find the failure modes on a bench rather than on a two-metre machine.

As technical lead I own the mechanical architecture and the decisions that follow from it: setting requirements, defining interfaces between sub-assemblies so people can work in parallel, and running design reviews where a choice gets made rather than deferred.

The practical part of the role is translation. Stakeholders describe a measurement problem; the electrical and controls sub-teams need motion envelopes, mounting interfaces and load cases. Holding that translation consistent — and writing down why a decision went the way it did — is most of what keeps a project of this size coherent.

The system is in development. Architecture and envelope are defined, mechanical design of the gantry and end effector is in progress, and subsystem prototyping is underway.

Next milestones: complete the single-axis repeatability prototype, close out the structural and alignment strategy at full span, and finalize the survey-meter interface for calibration access. Figures on this page are design targets; no repeatability, coverage or safety performance has been validated yet.