Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
A Six Sigma Black Belt Project
Transcript of A Six Sigma Black Belt Project
Quality problems were eating up most of their profits.
$140,000 scrap cost in
Define - Measure - Analyse - Improve - Control
Write the problem statement - what are the responses that need to be improved (these are the Y's).
Measure the response variable you are interested in improving in order to understand current process performance. You need to understand what is happening in the process.
Use data analysis techniques to understand which variables (the X's) are driving your response variable. Find the cause and effect relationships in the process.
We have measured a large number of samples and seen the histogram. Now lets look at the data plotted in time order - the order that the components were manufactured. This provides a view of the behaviour of the process.
Confirm and remove root causes.
Reduce process variability.
In this case, we learned that operators were overadjusting so we ended up hardwiring some controls and removing others.
Establish controls (spc, maintenance)
to ensure sustained improvements.
Reject rate = 115,000 dppm
2.7 sigma process
Reject rate = 17,500 dppm
3.6 sigma process
Brief overview of a Six Sigma Project
Dried powder mixture is dispensed
by a loss-in-weight feeder
Down a vibrating chute onto a wire
mesh along a moving conveyor
Plow acts as a bridge, limiting the amount
of powder that can pass through
Spreader distributes the powder evenly
across the width of the wire mesh
The large reels of laminate were then cut into lengths, processed further and then weighed.
Far too many of them failed.
That's bottom-line money.
You'd need an additional $1,000,000
in sales to generate that sort of profit.
With time-series data, always view the autocorrelation plot and a few lag plots.
These guys can help. I hear they're pretty good.
In this case, the third plot shows a relationship between items that are three units apart. So if an item is high in weight, the item three units away will also be high. And similarly, low weight follows low weight. This shouldn't happen with our process but the data clearly shows a correlation.
The most costly defect was high/low
weight of a specific component.
So the Y in Y=f(x) is weight.
After measuring a large sample of these components we make a histogram which shows us the distribution of weights. This one is heavily skewed on the high side.
One of the x's that is driving the Y (weight) is the behaviour of the laminator. This time series plot shows that the process experiences sudden and significant shifting.
On further review of the process, it was found that the operators were constantly adjusting the settings. If the laminate appeared thin, they would increase the speed of the feeder or slow down the belt or adjust the plow or... There were several variables and constant adjustment of these process inputs is what Deming calls tampering. Over-reaction to variation only increases variation. (Google the Deming Funnel if you want to see more on this idea).
The difference was remarkable. By removing the ability to overadjust, we eliminated a dominant source of variability, tampering.
The slugs we punched out provided 20 data points for every one data point we had before. The increased resolution in our view of the process revealed a striking behaviour. The peak-to-peak spacing was 12 seconds or 36 inches. Something in our process was repeating itself with that frequency.
The root cause was a worn belt that had a thin section which would allow more powder under the bridge once every 36 inches. Replacing the belt removed another large cause of variation in the process.
These two root causes allowed over a hundred thousand dollars back to the bottom line where it belongs.
Neither of these were identified in the brainstorming sessions.
Guessing doesn't work.
Problem solving requires specific skills.
Model Thinking - Applied Statistics - Data Analysis