DETROIT -- Stung by rising warranty costs, General Motors decided in the mid-1990s to pull design work for ignition and turn-signal switches from suppliers and put its own employees in charge. One of the first projects for the in-house team was the ignition switch for the Saturn Ion and Chevrolet Cobalt.
"We wanted to have control over the design," Ray DeGiorgio, the lead design engineer for the Ion and Cobalt ignition switch, said in an April 2013 deposition obtained by Automotive News. "So we brought them in-house."
That part has now been linked to at least 34 crashes and 12 deaths over the past decade. It's also at the center of a deepening mystery in the wake of GM's recall of 1.6 million 2003-07 vehicles fitted with the defective ignition switch:
Why did GM authorize a redesign of the part in 2006, eight years before the recall? And why was the change made so discreetly -- without a new part number -- that employees investigating complaints of Ions and Cobalts stalling didn't know about it until late last year?
These questions, among many that will be posed by lawmakers and federal safety regulators looking into GM's handling of the recall, have confounded some former GM engineers, who say the company's reports to regulators describe a sequence of events that was fundamentally at odds with standard operating procedure.
Not assigning the new part number would have been highly unusual, according to three people who worked as high-level GM engineers at the time. None of the engineers was involved in the handling of the ignition switch; all asked that their names not be used because of the sensitivity of the matter.
"Changing the fit, form or function of a part without making a part number change is a cardinal sin," said one of the engineers. "It would have been an extraordinary violation of internal processes."
Tracked with precision
The 2006 redesign was an important turning point. Not only did it fix the flaw and prevent possibly millions more cars from being affected, but the scant documentation of the change impeded GM's years-long search for clues.
GM now says the 2-inch-long switch, which DeGiorgio said the company began designing in 1997, didn't meet its specifications until the 2006 modification. Yet at the time GM didn't recall any vehicles that used the old switch.
GM said in a filing with the National Highway Traffic Safety Administration that the GM design engineer responsible for the Cobalt's ignition switch signed a document on April 26, 2006, approving the change, which was in response to a proposed redesign by a Delphi Mechatronics engineer to increase the torque in the switch with a new detent plunger and spring to prevent it from slipping.
Normally, the former GM engineers said, such component changes are tracked with precision -- down to the individual vehicle that was first to get the new part -- because the effects ripple through the company. Manufacturing plants must ensure that obsolete parts aren't used. Dealerships need to know they have the right replacement parts. And since component redesigns often involve cost changes, the purchasing department typically is involved.
Redesigning a part is not something done unilaterally by one engineer, the sources said, despite what is implied in GM's sparse account to NHTSA of the April 26, 2006, sign-off. Any alteration of a part at GM during that time, especially to a safety-critical component that had been flagged as potentially flawed, would have gone through a rigorous evaluation, the sources said.
It likely would have been vetted by a group of employees involved in the vehicle's development, called a change-review board, which would include other engineers -- possibly even an executive-level engineer -- as well as representatives from purchasing and finance, the sources said.
Ultimately, the engineer in charge of that component -- the design-responsible engineer, or DRE in GM parlance -- would sign off on the change. The engineer's superior, likely an engineering group manager, also would have to approve the change, the sources said.
"It's not like one or two guys can just lob things into the system," the second engineer said.
A GM employee would be responsible for assigning a new part number internally, two of the engineers said. But even if that didn't happen, a supplier isn't supposed to ship a redesigned part under an old identification number, the sources said.
"Failure to change a part number, at best, suggests sloppiness on both the part of the supplier and GM," the third engineer said.
Searching for explanation
By 2012, the lack of a distinct part identifier for the updated switch left GM engineers groping for an explanation of why the reports the company was getting of frontal crashes without airbag deployment involved only 2007 and earlier models. Finally, an outside expert hired by GM in April 2013 figured out that the switch had been changed for production during the 2007 model year.
The design change had been brought to GM's attention in a lawsuit filed by the estate of a Georgia woman who died when her Cobalt's ignition moved to "accessory" mode seconds before another car crashed into hers. Engineers and a mechanic hired by the victim's lawyer discovered that the ignition switches in Cobalts and Ions built after 2006 had as much as twice the torque of those in older models of those cars, meaning far more force was required to alter the position of the switch.
In late October of 2013, a month after GM settled that lawsuit, Delphi provided GM the records detailing the 2006 design change, according to the timeline GM filed with NHTSA. A committee of senior executives authorized a recall on Jan. 31 of this year.
GM says it now has located the April 2006 document authorizing the change to the ignition switch. It has not said who signed it. NHTSA has asked the company for the document.
In the April 2013 deposition, which was part of the Georgia lawsuit, DeGiorgio described himself as the "main responsible engineer for that project as it relates to the ignition switch." He insisted, repeatedly, that he had been unaware of the design modification.
"I was not aware of a detent plunger switch change. We certainly did not approve a detent plunger design change," DeGiorgio said at one point.
"So if any such change was made, it was made without your knowledge and authorization?" a GM lawyer later asked him.
"That is correct," DeGiorgio replied.
'I gave the car to my kid'
In his deposition, DeGiorgio -- whose LinkedIn profile indicates that he still works for GM and lists his current role as lead engineer for some push-button ignition switches -- dismissed the notion that GM would have made such a change quietly to avoid spending time or money testing the new design. He said the supplier would have been responsible for the revalidation process, which he said would have been necessary for such a change.
A GM spokesman declined to comment for this report and did not make DeGiorgio available for an interview.
Toward the end of the deposition, which was heavily redacted, DeGiorgio said he had bought a 2007 Cobalt "without any hesitation that this issue potentially existed" for his son to drive to college. "There's no way I would have done that -- I gave the car to my kid -- had I had any reservations," he said.
DeGiorgio said neither he nor his son had experienced any instances of the car stalling or the ignition shifting to "accessory" mode.
It's possible that his car contains one of the redesigned switches, even though it is included in the recall. GM told NHTSA that it believes Delphi began shipping the redesigned switch from a plant in Mexico "at some point during the 2007 model year."
DeGiorgio said in the deposition that GM had selected Eaton Corp. as the switch supplier for the Ion sometime before he took over the project in October 1999. Delphi, which was spun off by GM in May 1999, then took over production after completing its $300 million acquisition of Eaton's Vehicle Switch/Electronics Division in April 2001.
Delphi has not responded to calls seeking comment, but the company told NHTSA that it did not sell the recalled switch to any other automakers.
GM says the first documented incident of low detent plunger force in the switch occurred in 2001, on a preproduction version of the Ion, which was introduced in the 2003 model year. The same switch was later integrated into the Cobalt, which went on sale in 2004 as a 2005 model, and other vehicles that are part of the recall.