Driving quality with six sigma

Manufacturing Engineering, Feb 2003 by Olexa, Russ

Six Sigma

By designing parts using math-based software, Design for Six Sigma, and Six Sigma, GM has increased vehicle quality while lowering costs and improving its products' reliability and durability

When a transmission in one of its upscale vehicles began to shift erratically, GM knew it had a problem. But what caused it and how could they solve it quickly and efficiently? As it turned out, the problem involved more than a simple fix. Here's how GM solved it using Design for Six Sigma.

Six Sigma and Design for Six Sigma (DFSS) are two terms that GM uses in its product engineering programs. Statistical Engineering is GM's method to analyze an existing problem and bring a solution to it, using data and a standard process that is similar to Six Sigma. GM calls this type of work remediation-fixing problems that already exist. Design for Six Sigma is a problem-prevention tool used by product engineering to avoid the need for remediation efforts. In effect, they are eliminating problems before they occur, during the design phase. DFSS allows them to create products that will meet customer expectations under all operating conditions through robust designs that can be manufactured to the highest quality levels and are extremely reliable and durable in the customer's hands. GM used DFSS to solve the transmission problem that other QC methods couldn't.

GM Powertrain had to improve the sealing capability of an automatic-transmission aluminum die-cast channel plate that guides hydraulic fluid and controls gear shifting. It wasn't a matter of just adding a thicker gasket and increasing the bolt torque. "Sometimes we have a product problem where we can't separate out the primary cause of the failure," says Executive Director Powertrain/Vehicle Integration Shawn Burns.

"We were experiencing what's called fourth clutch-band distress," says Burns. "There are bands and clutches in a transmission that allow it to shift from one gear to another Distress means band wear or overheating." This problem was a warranty issue and Burns said they are constantly attacking their top warranty problems using DFSS and Statistical Engineering. "This was a remediation project, meaning it was a field problem that we wanted to solve with DFSS. Normally, we attack new designs with DFSS to make sure they are right the first time before they get into the field. We went with DFSS on this one, because our statistical engineering methodology couldn't identify the corrective actions needed in the manufacturing process to make the part more robust. It was not just an issue of improving process control. We needed to alter some multiple interaction processes," notes Burns.

The channel plate was warping and not allowing a proper seal. GM used a combination of DFSS and math-- based software techniques (first principal equations that represent how a manufacturing process should perform) to study the problem. Using these techniques, they created a functional-shape specification for the part, and a measurement system to see how a process change would affect the part's dimensions and performance.

"We did this with math using off-the-shelf finite element software in conjunction with internally developed codes and methods.

These methods allow us to make process predictions with software instead of the traditional trial and error methods using hardware. Time and money are the true savings," says senior project engineer Dan White. GM also used hardware techniques for product validation such as pressure-sensing paper in between the channel plate and its mating components. The transmission bolts are torqued down, and the sensing paper measures local contact pressure. Through these methods, they validated the accuracy of the math-based model.

For the project, they created a structure diagram that shows the physical connections of the part. They also used a scorecard that names all the parts and systems that are in the structure diagram, and it explains the critical requirements that each part in the system must meet to achieve the quality target. "We work the process for all types of requirements relating to supplied parts, internal parts, performance, and reliability," adds Burns. "In this situation, the Critical to Quality (CTQs) categories were in the performance and manufacturing areas, like how well does the transmission shift, and will the channel plate seal."

They also do a functional diagram to identify the transfer functions. For instance, GM needed the transmission to be reliable and dependable with good shift quality, all of which were customer CTQs. Then the team cascades these CTQs down to the transmission, clutch system, and eventually to the channel plate. "Our engineers then make a decision on what's important. Part of this decision is based on data and others on empirical relationships. In this project, we broke it down to the smallest components that could affect shift quality or sealing. Then the math-modeling equations were exercised to predict the effect of the proposed process changes on the CTQs," adds Burns.