Tag: manufacturing

Before a new production line is cleared for volume output, it typically looks like a success story: parts come off correctly, fixtures hold, sensors read green, and inspection steps pass. The process has demonstrated it can build the thing it was designed to build. But what nobody has established is what the line would do with a part built wrong.

Neelamegam Subramanian, a technical program manager for global fleet electrification and charging infrastructure at a major logistics and technology company, believes that gap is where most preventable defects are born. “Before a new production line is trusted to build products at volume, somebody has to deliberately try to break it,” he says. He adds that sensors and inspection steps may look complete on paper, but “the only real way to know what they’ll miss is to build a part wrong on purpose.”

Subramanian came to this approach through quality engineering, including several years at an energy storage and electric vehicle manufacturer, where he handled casting and machining quality for a high-volume vehicle line and later led launch quality for a large-scale battery energy storage product. He earned his master’s in industrial and systems engineering at the University of Southern California.

The practice he is known for started at line qualification. Before a new line was cleared for volume, he designed and ran fault-injection studies, introducing controlled defects into the process on purpose: misaligned welds, missing fasteners, out-of-spec fits. Then he recorded which failure modes the line caught and which passed through untouched.

He is direct about why this is unusual. “Most quality validation at that stage focuses on whether the process can build a good part, not on actively trying to build a bad one and seeing what gets missed.” Deliberately introducing failure modes into a line about to enter production is also, as he puts it, a harder sell operationally, and there was no established playbook for it on that program.

The studies found detection blind spots that were invisible from the process design alone. Those gaps were closed with revised inspection steps and control-plan updates before the line ever ran at full volume, which meant the failure modes never had the chance to reach a customer. The timing is the whole argument: a blind spot found at line qualification costs a revised inspection step, whereas the same blind spot found after launch costs sorting, rework, or a defect that reaches the field.

That instinct carries into how he handles trouble once a line is already running. During the ramp of a new energy storage line, a weld-alignment defect appeared on a structural bracket, stopping the part from seating in its assembly slot. It was not a known failure mode, so nothing in the process documentation or inspection plan anticipated it, and there was no fix to reach for. Stopping the line during a critical ramp and continuing to build parts that might need rework were both bad options. Instead he built a third one: a dry-fit inspection using a dummy part to confirm fit before installation, which kept production moving while a permanent correction was engineered and validated in parallel. Roughly 500 units went through assembly that way without rework. He estimates the avoided rework at more than $1 million, and is careful to call that a rough figure rather than an audited one. The containment method was written up so the same defect would not have to be diagnosed from scratch if it appeared again.

“How you contain a problem matters just as much as how fast you find the root cause,” he says. A good interim fix removes the choice between stopping the line and shipping something broken.

His wider record runs along the same lines. In a quality manager role at a casting supplier site, he held defect rates under 15 parts per million, and in another role sustained 200 consecutive days without a major customer complaint. Root-cause work on a run of poor incoming quality produced lasting improvement across roughly two dozen supplier-made parts, and earlier in his career he helped cut scrap costs by more than $250,000 a year at a battery manufacturer. In two previous roles he led teams of quality engineers and technicians.

That history is what he now applies to suppliers rather than lines. Judging whether a supplier can deliver, he says, means testing stated capability instead of taking it at face value, the same way he tested a line’s detection systems rather than trusting the design. He’s also been observing patterns he kept running into on the floor, including Foreign Object Debris (FOD) risk, weld quality variability, part-level traceability, and turning them into formal frameworks he’s now preparing for journal submission. His take on traceability cuts against the usual instinct: when a manufacturer and a supplier are looking at the same part-level data during a failure investigation, “it stops being about who’s to blame and starts being about actually fixing the problem.”

The misconception he’s most eager to retire is treating a quality issue as a one-off. The fix that actually holds, he says, “almost always means going deep enough to change how the part is actually made or inspected, not just sorting out the bad ones as they come in.”