How to qualify a second-source motor

A second source exists to remove a single point of failure from a supply chain. It only does that if the second part can be substituted without a change notice to the customer, without a mechanism change, and without a different field failure rate. Qualification is the work of establishing all three, and it is the only thing that separates a second source from a second supplier of an unqualified risk.

The structure of the work is the same as any motor change: fit, performance, life, and now a fourth claim that a single-source change does not carry, which is that the two sources stay equivalent over time.

Define equivalence before you test anything

Write down what equivalent means for this part, in measurable terms, before samples arrive. Otherwise the qualification becomes a negotiation about which differences matter, conducted after the differences are already known.

• Interchangeable dimensionally, with the same mounting datum, pilot, shaft and lead-out. State which dimensions are controlled and which are reference.
• Within a stated band of the incumbent on no-load speed, no-load current, stall current, torque constant and, for a gearmotor, backlash and drag.
• Equal or better on the life-limiting mechanism identified for this application, demonstrated at the application duty cycle.
• Manufactured under change control that requires notice before any change to winding, magnet, brush system, bearings, gear material, lubricant, or manufacturing location.

Characterize both sources, not just the new one

The most common qualification error is testing the candidate against the incumbent's data sheet. The data sheet is a nominal. The population you actually receive has a center and a spread, and both matter. Characterize a sample of production incumbent parts at the same time and on the same equipment as the candidates, or the comparison contains the measurement system's differences as well as the parts'.

How many units the life test needs

If a life test runs to completion with zero failures, the result is not "it works". It is a statistical bound, and the bound depends entirely on how many units ran. The relationship is exact for a pass or fail test with no failures.

Two numbers are worth memorizing. Twenty two units through a full-duration test with zero failures demonstrates 90 percent reliability at 90 percent confidence. Forty five units demonstrates 95 percent. Five units, which is a common practical sample, demonstrates about 63 percent reliability at 90 percent confidence, which is not a useful statement about a production part.

Qualify the process, not only the parts

The fourth claim, that the two sources stay equivalent, is closed by process controls rather than by measurements on parts. This is the part of a qualification that gets deferred and then never done, and it is what determines whether the second source is still equivalent in three years.

1. A control plan that names the characteristics under control and the method and frequency of control for each.
2. Capability evidence on the characteristics that matter, from a stable process, not from a sorted sample.
3. A change notification agreement covering material, process, tooling, sub-tier supplier and manufacturing location, with a stated notice period before shipment of changed parts.
4. Lot traceability that allows a field failure to be tied back to a build lot, on both sources.
5. A first-article process for any changed revision, on the same terms as the original qualification.
6. An agreed re-qualification trigger list, so a change does not have to be argued about when it happens.

Running two sources

Once both are qualified, the mechanism has to keep working with either, at any point in each source's tolerance range. That is a different requirement from working with each source's nominal.

• Build the worst-case combinations deliberately: the slowest candidate against the tightest mechanism, the highest drag against the lowest supply voltage.
• Keep both sources in production at a meaningful rate. A source that has not shipped in two years is not a second source, it is a lapsed qualification.
• Monitor the same parameters on incoming material from both, on the same equipment, so drift in either is visible before it becomes a field trend.
• Keep the comparison data. When a field problem appears, the incoming characterization history is what distinguishes a source difference from a design margin problem.

When time is short, test in risk order

A full qualification is weeks of work, and it is often started because a supply problem is already happening. When the schedule cannot carry the whole program, run it in the order that retires the most risk per day rather than in the order the document lists.

1. Fit first. It is hours of work, it is unambiguous, and a fit failure ends the evaluation before anything expensive starts.
2. Then whatever is structurally different. If the candidate has a different bearing type, a different brush system or a different gear material, that is the claim most likely to fail, and it is worth starting the long test on it immediately even while other work continues.
3. Then performance in the mechanism, across the mechanism's own tolerance range rather than on one good unit.
4. Then the life test, started as early as possible because it is the long pole and its duration cannot be compressed without changing what it measures.
5. Process controls in parallel throughout. They are documentation work and they do not compete with bench time.

Starting the life test early is the single most useful scheduling decision, and it is the one most often deferred until the other results are in. A life test started in week one and confirmed in week eight is a qualification. The same test started in week eight is a risk acceptance with a test running behind it.