nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

DSC (Differential Scanning Calorimetry) - is JMP capable of doing this or helps with this?

martindemel

Hi,

a customer producing different carbon fiber products is working on data from a DSC (Differential scanning calorimetry more info to that at the bottom)

The customer gets data for the temperature over time, where within the time there are set different reactions (e.g. oxidation) to see the effect it has on the material, base line, stability of the polymer and more. She wants to analyse e.g. the heat rate variation over each reaction and need to differentiate the reaction effects from the temperature curve.

Has anyone ever have done such kind of work in the past or worked with a customer on such things? How could JMP help with these kind of analysis?

Currently she uses an Excel solver , then several methods, like simpson, kissenger. But Excel is not stable in the results, means results differ where they shouldn't. Identifying the peaks as well knowing that JMP has methods is one thing, but I need some guidance on how one would proceed with such an analysis within JMP.

Any hints are highly welcome, best before Monday afternoon. Thanks a lot

Martin

Differential scanning calorimetry or DSC is a thermoanalytical technique in which the difference in the amount of heat required to increase the temperature of a sample and reference is measured as a function of temperature. Both the sample and reference are maintained at nearly the same temperature throughout the experiment. Generally, the temperature program for a DSC analysis is designed such that the sample holder temperature increases linearly as a function of time. The reference sample should have a well-defined heat capacity over the range of temperatures to be scanned.

more here: Differential scanning calorimetry - Wikipedia, the free encyclopedia

Accepted answers

Bill_Worley

Martin,

Life Distribution in JMP 12 allows you to look at different combinations of different distributions. Use Fit Mixture under the Life Distribution hot spot.

Bill

All comments

Bill_Worley

Hi Martin,

Someone I worked with at P&G who took DSC data from various compounds/materials and was able to model/predict glass transition and melt point. I believe he used PCA to get the factors and then used NN with the principal components as predictors. I will ask if he is willing to share his technique and let you know.

Best,

Bill

martindemel

That would be awesome!

erich_gundlach

Martin,

I did an extensive amount of work using DSC, isothermal microcalorimetry, and other thermoanalytical techniques in my previous work as a chemist. While JMP doesn't have much to offer over the software provided by instrument manufacturers with respect to "marking" the thermograms (i.e. peak integration, measuring step transitions, marking and subtracting baselines, etc.), JMP far exceeds the capabilities of that same software away when it comes to modeling that data after doing the initial "marking." I used JMP along with thermoanalytical data to generate models for determining purity and amorphous content and for fitting non-linear reaction rate models, among other things.

Erich

martindemel

Hi Eric,

that is really helpful. I'm looking Forward to what the customer will be saying.

I'll Keep you updated here.

Thanks a lot,

Martin

martindemel

Hi Bill and Eric and others,

I have got new insights from the customer. She does not have a machine to identify the Peaks, she just gets the data which may look like the curve on the left (b) for the heat flow. The left part of the curve is maximized on the right (Bild 8.). And what she wants to identify is, in this case the curves for 4 reactions (she has three reactions) based on different distributions, e.g. Gauß, Lorenz and/or Weibull. In her case she said, usually normal distributions do not reflect the reaction influence on the heat flow, Lorenz and Weibull do a better job. In the (German) paper this is named dividing a C-NMR spectrum by Peak Separation using a fit program.

So she's already stucked at the beginning identifying the peaks for the reaction influence functions/distributions rather the peaks of the heat flow itself.

In JMP I could look for single or normal mixtures distributions. But I do not see a way to detect e.g. what mixture of distributions could lead to the overall curve. Is there any and at best easy way to play around with different Distribution combinations? Or could I use one f the modeling platforms to identify those fits?

Thanks, Martin

Bill_Worley

Hi Martin,

Please comments below from Cy Wegman at P&G regarding his experience analyzing DSC data.

The most sure way to model the DSC is to use the summary
measures you get from the DSC. Everyone accepts those and understands them.
Ideally you model them simultaneously since they are from the same measure. You
can do this in Neural models but also in JMP 12 with the new selection
technique in the least squares platform you select model parameters at the
simultaneously.

Modeling the entire DSC curve may be done via a functional data
analysis approach or more simply with PLS. The problem with this approach is
the heat flow numbers change with the sample size so the curves are shifting up
and down. So to help alleviate this I normalize them to a constant starting
point. This still does not fix it perfectly though. If you were really going to
try to use this approach in earnest you should use care in the sample size and
repeat the DSC measure several times to average out the sampling error. As far
as I know, no one was willing to do that here.

For the PLS approach you just use the Fit Model platform and
profile is the response. Each position is its own column. If the DSC curves
have different number of points you can ‘normalize’ them by fitting each
individually with a neural model and then make a common scale and use the
neural model to give you the y for each scale point.

Bill_Worley

Hi Martin,

Please the response I received from Cy Wegman at P&G regarding using JMP to model DSC data.

Bill_Worley

Martin,

Life Distribution in JMP 12 allows you to look at different combinations of different distributions. Use Fit Mixture under the Life Distribution hot spot.

Bill

martindemel

Hi Bill,

Thanks, although it’s true you could use fit mixture to fit the whole curve you will not get the shapes of each of the mixed distributions, right? But may I could use the estimates for each of these and plot my own curves in an overlay plot together with the whole data.

I will follow up this thought.

Martin

Martin Demel

JMP Systems Engineer ▪ JMP Sales ▪ Germany, Austria, Switzerland

Tel: + 49 89 15929 315 ▪ Fax: + 49 6221 415 44 6115

Mobile: + 49 172 6340 872 ▪ Martin.Demel@jmp.com<mailto:Martin.Demel@jmp.com> ▪ www.jmp.com/de<http://www.jmp.com/de>

Adresse: Nördliche Auffahrtsallee 22 ▪ 80638 München

martindemel

Hi Bill,

thanks a lot, this is helpful to understand the further process. I used the Fit Mixture Feature in Life Distribution and it seems to do the Job. I will have a final call this week with the customer to get his Feedback if that's enough and what she wants.

I've learned a lot, thanks!

Martin