Flame Treatment Process - Warning: Engineering Content
Moderators: Head Monkey, kelvin, bigKam, skidesmond, chrismp
Flame Treatment Process - Warning: Engineering Content
So in another attempt to include a little engineering into ski building, an experiment was developed to analyze the effects of flame treatment.
For you engineerds, the experiment is a central composite design with three center points. For the non engineers, basically all you need to know is that this design can measure the amount of variance in a process, so the results should have statistical significance - even from a small sample size.
I used a contact angle goniometer to measure water contact angles, which are an indicator of surface energy. Low contact angles = high surface energy = better bonding. For a little perspective, polycarbonate has a contact angle of 78*, and ABS has 71* - both are considered to be fairly bondable polymers.
Regression Statistics
Multiple R 0.938567996
R Square 0.880909883
Adjusted R Square 0.761819766
Standard Error 11.71344554
Observations 11
ANOVA
df SS MS F Significance F
Regression 5 5074.521422 1014.904284 7.397002412 0.023275481
Residual 5 686.0240324 137.2048065
Total 10 5760.545455
Coefficients Standard Error t Stat P-value
Intercept 65.06119535 6.761841209 9.621816505 0.000205602
Factor A - Distance -16.46464646 4.162191634 -3.955763672 0.010788402
Factor B - Time 1.54040404 4.162191634 0.370094454 0.726471486
A^2 -15.48776579 4.9963464 -3.099818257 0.026858716
B^2 -19.82450048 4.9963464 -3.967799447 0.010659545
AB -6.5 5.856722771 -1.109835697 0.317573497
Sorry that's so butchered... not sure of the best way to plug in spreadsheet data.
Anyway, the distance from the flame is significant linearly and exponentially (has curvature), and time is significant with curvature. Also, 76% of the data is explained by this model, indicating these two factors are the key elements in flame treatment (duh...)
Now we can use this data to generate a surface plot, which will tell us the optimum value for each factor.
See that tiny little spec of purple on the left? That's our optimum - 4" from torch tip and 1.7 to 1.9 seconds of treatment time.(per 1/2x1/2" section) This region produces a water contact angle of 46*. Now this probably isn't a surprise to all of you, but now we know those values are repeatable and will result in a good bond!
Thanks for reading!
For you engineerds, the experiment is a central composite design with three center points. For the non engineers, basically all you need to know is that this design can measure the amount of variance in a process, so the results should have statistical significance - even from a small sample size.
I used a contact angle goniometer to measure water contact angles, which are an indicator of surface energy. Low contact angles = high surface energy = better bonding. For a little perspective, polycarbonate has a contact angle of 78*, and ABS has 71* - both are considered to be fairly bondable polymers.
Regression Statistics
Multiple R 0.938567996
R Square 0.880909883
Adjusted R Square 0.761819766
Standard Error 11.71344554
Observations 11
ANOVA
df SS MS F Significance F
Regression 5 5074.521422 1014.904284 7.397002412 0.023275481
Residual 5 686.0240324 137.2048065
Total 10 5760.545455
Coefficients Standard Error t Stat P-value
Intercept 65.06119535 6.761841209 9.621816505 0.000205602
Factor A - Distance -16.46464646 4.162191634 -3.955763672 0.010788402
Factor B - Time 1.54040404 4.162191634 0.370094454 0.726471486
A^2 -15.48776579 4.9963464 -3.099818257 0.026858716
B^2 -19.82450048 4.9963464 -3.967799447 0.010659545
AB -6.5 5.856722771 -1.109835697 0.317573497
Sorry that's so butchered... not sure of the best way to plug in spreadsheet data.
Anyway, the distance from the flame is significant linearly and exponentially (has curvature), and time is significant with curvature. Also, 76% of the data is explained by this model, indicating these two factors are the key elements in flame treatment (duh...)
Now we can use this data to generate a surface plot, which will tell us the optimum value for each factor.
See that tiny little spec of purple on the left? That's our optimum - 4" from torch tip and 1.7 to 1.9 seconds of treatment time.(per 1/2x1/2" section) This region produces a water contact angle of 46*. Now this probably isn't a surprise to all of you, but now we know those values are repeatable and will result in a good bond!
Thanks for reading!
-
- Posts: 2337
- Joined: Tue Apr 07, 2009 3:26 pm
- Location: Western Mass, USA
- Contact:
Nope, and all those are good reasons that only 76% of the data is explained in the model. What this is showing is that distance from torch tip and treatment time are the most important.chrismp wrote:great little article! i have no clue what all the numbers mean, but i get the essence
one question though: i always thought the bonding is influenced by a couple of other variables like type of flame, flame size, oxygen available in the room, etc.
are those things neglectable?
But this reminds me, I used a propane gas torch - if you use a different type of torch you will have different results.
Also, the material was UHMWPE - I couldn't recommend these times/distances for other materials as I haven't tested other materials. I referenced ABS and PC as they are fairly readily bondable polymers off the shelf.
Damon
You could decrease the factor levels to decrease error, but the nice thing about a CCD is you can leave the levels fairly broad to hone in on an optimum.Alex13 wrote:Great info, cheers.
Have you tried closer contact points? Is there a chance to increase the error range for flame treating (with regard to time) by decreasing the distance below 4"?
- MontuckyMadman
- Posts: 2395
- Joined: Fri Jun 20, 2008 9:41 pm
4"? The flame doesn't even come close to touching not even with a straight flame I have used a flame spreader, as this was suggested.
Tap plastics suggests a different application method.
I don't understand closer could be better? How come that's not on the graph?
DBS, you have done this a ton does the enginerd bs stand up to actual application?
Tap plastics suggests a different application method.
I don't understand closer could be better? How come that's not on the graph?
DBS, you have done this a ton does the enginerd bs stand up to actual application?
Well you get too close and you start cooking it, which actually increases the contact angle as it smooths the surface. The reason 4" is the minimum is because of the way the design is structured. The experiment sees samples at 2.6", but can't include it in the design space because the function doesn't know what goes on after that value. So the values known between 4" and 2.6" give it the curvature required to show a peak/valley/whatever.
Now, to do a full scientific experiment I should go back and find that peak. But knowing that area results in a decrease of 80-100 degrees (water contact angle), and UHMWPE has a natural contact angle of 120*, that area is the best I can hope for giving me a theoretical value of 20-40*. The best we saw in the experiment was 46.
4" on my torch was at the end of the flame.
Now, to do a full scientific experiment I should go back and find that peak. But knowing that area results in a decrease of 80-100 degrees (water contact angle), and UHMWPE has a natural contact angle of 120*, that area is the best I can hope for giving me a theoretical value of 20-40*. The best we saw in the experiment was 46.
4" on my torch was at the end of the flame.
- MontuckyMadman
- Posts: 2395
- Joined: Fri Jun 20, 2008 9:41 pm
I guess in the dark 4" is at the end on a pretty sweet propane torch, and you want it in the oxidation zone.
I don't understand your regression analysis and that is my own downfall.
I know melty is way to hot.
Good data I guess I will back off and slow down on the flame.
I'm still using s spreader though.
I don't understand your regression analysis and that is my own downfall.
I know melty is way to hot.
Good data I guess I will back off and slow down on the flame.
I'm still using s spreader though.
-
- Posts: 1354
- Joined: Mon Sep 17, 2007 7:37 pm
When I read it this morning, I thought it sounded about right.MontuckyMadman wrote:4"? The flame doesn't even come close to touching not even with a straight flame I have used a flame spreader, as this was suggested.
Tap plastics suggests a different application method.
I don't understand closer could be better? How come that's not on the graph?
DBS, you have done this a ton does the enginerd bs stand up to actual application?