|
This thread is carried over and continued in the Current Elsmar Cove Forums 
|
|
|
This thread is carried over and continued in the Current Elsmar Cove Forums
|
The New Elsmar Cove Forums
|
 The New Elsmar Cove Forums
 Statistical Techniques and 6 Sigma
 Control chart application for post heat treat processes
|
 UBBFriend: Email This Page to Someone! |
next newest topic | next oldest topic |
| Author | Topic: Control chart application for post heat treat processes |
|
Mark W Lurker (<10 Posts) Posts: 6 |
posted 14 July 1999 01:52 PM
  
I have struggled with trying to find a method/solution/tool...or something to analyze a process in which dimensional product characteristics change after the parts are heat treated. Our company performs induction heat treat internally where only 2 parts are heat treated at a time, and we also send batches of parts out to subcontractors for carbonitride where thousands of parts might be furnace heat treated simultaneously. For some critical characteristics we have Xbar-R charts in use at the machining operation, and the process demonstrates a statistically controlled process. The problem arises when variations in material chemistry, quench rate, etc. cause these dimensions to both "grow" and "shrink" in sometimes significant amounts. For our internal heat treating, we perform many different studies to try to control the process variables so that we can try to predict where to center the machining process in order to meet the specifications after heat treat. Then periodically we take samples on these dimensions as the heat treating is performed, and chart them using another Xbar-R chart to determine our capability indexes and whether the process is still in statistical control. We have seen some success and many frustrations with this method. The problem is further complicated when we must try to position the machining process for large batches of parts which go to subcontractors for heat treat processing. One of the major obstacles is trying to develop a way of determining statistical control for batches of parts in which we are not running. We must rely on incoming sampling from these lots to try to determine compliance with the specifications, but determining statistical control and capability indexes from these samples doesn't quite make sense to me. Mainly because the sample we take is from a very large lot, and doesn't seem to be much different than if our customer were to sample our shipments, and try to determine if our process was in control or not. My primary question I guess is: Are there better tools available for this type of analysis or maybe a better application of the control chart to determine statistical control. Any insight from anyone who has had similar experiences would be greatly appeciated. IP: Logged |
|
Don Winton Forum Contributor Posts: 498 |
posted 15 July 1999 01:34 AM
The problem and frustrations you are experiencing are to be expected. You are monitoring the outputs to try to control the inputs. This will not work. The answers you seek may be beyond the capability of the control chart. You may require advanced statistical techniques.
quote: It appears that you have several variables, but concentrate on three for now. One, article temperature. Two, quench temperature. Three, quench time. I believe that you will find that, if these three are looked at first, variation in material may not be significant. Perhaps not. What you need is someone who can set up a designed experiment that considers these three variables first (after that, look at the material). Once you have determined how these variables affect your dimensions, then process control techniques can be utilized to control the inputs. Alternatively, put controls on the three variables I mentioned above. Control article temperature to 'X' degrees, +/- range. Control quench temperature to 'X' degrees +/- range. Same for quench time. Place controls on these and use control charts for these inputs. Drawing on my limited past experience, these controls may go something like this: Place article in oven at 'X' degrees for one hour +/- 10 minutes. Remove articles and place in {Type of here. Oil, water, etc.} quench of 'X' degrees +/- 5 degrees. Leave articles in quench for 5 minutes +/- 1 minute. I believe you will find that when you place controls on the inputs, the outputs will become more predictable. I would still advise the advanced statistical technique if you want to know the ideal inputs. I can help, if you like, but would need more details. Hope this helps. Regards, IP: Logged |
All times are Eastern Standard Time (USA) | next newest topic | next oldest topic |
  |
Hop to: |
Your Input Into These Forums Is Appreciated! Thanks!
Main Site Search 
