The Pain Point: The Prototype Was Perfect

Every programme has the same moment. Engineering receives five machined samples; they are beautiful, they measure well, and the design is signed off. Six months later the production batch arrives and the assembly no longer fits. Holes are 0.02 mm off, wall thicknesses vary, one part has visible tool marks on a datum face, and the anodizing colour differs between panels.

The prototype was not a lie. It was made on a different machine, in a different setup, by a different operator, possibly with stock from a different mill lot, and inspected by the engineer who designed it. None of that survives the transfer to volume.

Sample-to-serial consistency is the single most under-specified requirement in the aluminum parts supply chain. Customers ask for a tolerance; they rarely ask for evidence that the tolerance will be reproduced in month six.

The Principle: Variation Lives in the Process, Not the Part

Setup variation. Every additional fixturing change adds a datum shift. A part made in one setup has a different error signature than the same part made in four.

Stock variation. Rolled plate from different lots carries different residual stress and slightly different chemistry. Two plates machined identically can relax differently.

Tool wear. Aluminum is abrasive in its own way, particularly the silicon-containing alloys. A worn tool raises cutting temperature and alters surface integrity.

Thermal and dimensional drift. Machines move through the day. A shop that does not record inspection temperature will produce shifting results across a batch.

Finishing variation. Anodizing bath chemistry, rack position and dye lot all affect thickness and colour. These are process variables in the coating line, and they are the most common cause of "identical" parts not matching.

The chemistry is relevant here as well. Aluminum's eagerness to bond with oxygen means every process step leaves a different surface state: a freshly cut surface, a blasted surface, a chemically brightened surface and an anodized surface are four different interfaces. Each must be defined, not assumed.

There is also a statistical dimension that is frequently misunderstood. Conformance and consistency are not the same property. A process running at the centre of a tolerance band with small spread will produce identical parts. A process running at the edge of the band with small spread will also produce identical parts - identical, and all marginally out of specification after the next drift. A process with a wide spread can pass a sample inspection and still deliver a batch in which one part in twenty is out. Only variables data over a run of parts, not a sample inspection, distinguishes these three cases.

Equipment Classification: Consistency Risk by Part Type

Gauge blocks and reference standards. Highest sensitivity; any variation is a metrology failure.

Fixture plates and locating components. Hole position consistency governs whether a part can be loaded repeatedly in the same place.

Optical and sensor housings. Coat thickness and colour consistency matter as much as dimension.

Frame and structural components. Lower dimensional sensitivity but high cosmetic and assembly-fit sensitivity.

Turned inserts, pins and bushings with wear surfaces. Consistency of surface finish and coating depth determines life.

Practical Guide: Writing a Specification That Holds

Define the drawing as a process document. State alloy and temper by standard designation, stock form, stress relief condition, and the process sequence for critical features.

Use GD&T properly. Position tolerances referenced to stable datums, profile tolerances on contoured surfaces, and a clearly identified primary datum that will not be machined away later.

Require a first article inspection report with variables data. Attribute data hides trends. Variables data shows whether a process is centred or merely inside the limits.

Ask for the capability, not just the conformance. A Cpk of 1.33 on a critical dimension is a far better predictor of batch consistency than a single inspection sheet.

Freeze the process between sample and production. If the sample was made on a five-axis machine and production moves to a three-axis line with four setups, expect a different part.

Control finishing separately. Specify coating thickness range, colour reference, masking scheme and the inspection method after coating. Visit the coating line once; it explains more than a certificate.

Define traceability. Material certificates, heat or lot numbers, and inspection records tied to serial numbers. This is standard for ISO 9001:2015 systems but is often delivered as a generic certificate rather than part-specific evidence.

Field Experience: The Batch That Fit Nothing

A metrology OEM qualified a supplier on five 6061-T6 mounting brackets for a portable measuring arm. The samples were machined from stress-relieved plate, inspected on a CMM, and approved. The first production order of 300 pieces arrived with brackets that would not seat: a 0.03 mm bow across the mounting face meant the arm base rocked under load.

The supplier had switched stock from T651 plate to standard T6 to reduce lead time, and had moved the job to a machine with a different fixture arrangement. Neither change was communicated, because the purchase order only specified alloy, dimensions and finish. Every individual dimension was within tolerance; the assembly still failed.

The corrective action was structural rather than technical. The drawing was revised to specify T651 stock and a machining sequence. A control plan was agreed, including a flatness check after coating on 10 % of the batch. The supplier's process was frozen for the duration of the programme, and any change required written notification. Scrap fell from 22 % on the second batch to under 2 % on the fourth.

The expensive part was not the rework. It was the two months in which the customer's assembly line could not build.

One detail from that programme is worth keeping. The supplier had not been careless; it had optimised for lead time, which was the only metric the buyer had communicated. Standard T6 plate was available in three days, T651 in fifteen. Nobody had ever explained that the flatness requirement depended on the temper. Once the requirement was explained in those terms, the supplier proposed the correct stock itself and absorbed part of the cost of the change to keep the account. Communication of intent, not just of numbers, is what turned the relationship around.

Summary

Consistency in aluminum parts for precision measuring equipment is a product of process control, not of inspection. A supplier that can describe its stock specification, its setup count, its stress-relief sequence, its coating parameters and its capability data is a supplier that will deliver the same part in month twelve as in week one. A supplier that can only offer a dimensional report is offering a snapshot, and snapshots are what cause the second-batch failure.

Outlook

Procurement in this sector is shifting from price-per-part to total cost of variation. More OEMs are asking for control plans, capability studies and change-notification clauses as standard terms, and are auditing the finishing line as well as the machine shop. In parallel, digital traceability - a serial number that links material lot, machine, operator, program revision and inspection result - is moving from a premium option to an expectation. Suppliers who invest in that infrastructure will find themselves qualified for programmes that job shops cannot enter.