A recall for faulty ignition switches that could impact as many as 2.6 million cars made by General Motors has captured the headlines for the last few months, with the number of deaths, the years of accidents, the perceived inaction by the NHTSA, the first woman CEO, the relatively inexpensive fix (based on some news reports), and the staggering size of the recall, all contributing to the media frenzy.
But hidden deep in the news reports, is a possible engineering change control mistake, which if true, will significantly increase the scope of the recall, and will compound the financial impact of this problem for GM. For those of us that work in manufacturing, this is a learning moment, with an important lesson to share with our internal teams, as well as with our suppliers.
The engineering change control process requires that every time a design change results in a change to a component’s Form (shape, weight et.), Fit (interface with other components), or Function (reliability, serviceability, operating conditions), the component must be issued a new part number. This is required, because the new part will no longer be directly interchangeable with the old part.
Imagine walking up to a bin in a warehouse that has both the old components and the new components. If you reach in and pull out a part at random, you should be able to use it without having to worry about whether it’s an old part or a new part, because nothing important will have changed.
Media reports suggest that in the case of the faulty ignition switch, the supplier made a change to a spring to increase the force required to move the key in the ignition from ignition-mode to accessory-mode or off-mode. This would be classified as a change in “Function”. But, instead of issuing a new part number, the supplier issued a new part revision.
In most companies, part revisions are used to document minor changes that do not impact Form, Fit or Function. As an example, a correction to a document or drawing that does not impact the Form, Fit, or Function of a part will result in a new part revision. Since the old and the new versions of the part are interchangeable, two or more different revisions will typically be comingled in the same warehouse bin. As parts are pulled from the bin for production or for spares, a customer could receive an old part or a new part, and will not care about the difference.
Unfortunately, in the case of the faulty ignition switch, the new design was meant to be an improvement over the old design, but because part numbers were not changed, the old and the new parts could potentially be in the same bin in the warehouse.
There is now no easy way of knowing which cars received the old design as a spare, and which car received the new design as a spare. As part of the recall, GM will need to replace all parts made under that part number, even if some of them were made under the new design.
This engineering change control mistake is fairly common across industries. As an example, in the semiconductor capital equipment industry, changes to a component could result in on-wafer process variation. Many customers, including Intel, therefore insist on a stringent change control process (called Copy Exact!) which requires a detailed review and validation of changes to critical components before they are cut into production.
It’s very important for product manufacturers to recognize that outsourcing does not make problems go away. Therefore, even if they outsource a significant amount of manufacturing and design activities, they still need to ensure robust change control processes across the supply chain. They are closest to the customer, and in the best position to understand the impact of a change by a supplier. More importantly, at the end of the day, the product maker, is responsible for the safety, reliability, and quality of your products.