Saturday, 22 August 2015

W/c 17th August: Curves, KATs and more curves

This image pretty much sums up my week. For the most part, I've stared at LabVIEW graphs of curves being fitted: and that's about it. There are good reasons for this, though I won’t give out too much detail, since we'll need to hold that back for the publications. And a lot of it has been debugging code and rethinking models.

I had a particularly nasty bug where LABVIEW fitted a lovely curve, but returned a denominator which would always be zero in the output spreadsheet. I went through the code time and again. It all seemed right. When I tried it on one file at a time (writing output to the screen) it worked fine. Process a batch in one go and the summary spreadsheet returned the same impissible zeroes. Loads of them. Not all the time, but a lot. Yet some results matched perfectly with the individual runs.

And the solution, of course, was trivial: I'd set LabVIEW to write the spreadsheet to 3 decimal places. So 0.0006 showed up as 0.001 (close enough), but 0.0004 showed up as 0 - not 0.000, just 0. One character replaced and the problem was solved.

It was worth it, though - it all seems to be coming together, and after weeks of investigating and rethinking, we are (to quote my colleague Mark Mon-Williams) "on the verge of something big".

It was a week where I did little else. But just as you have some weeks where you do much but get little done, the payoff is that you get some weeks where you do little, but get a lot done! So, revised electronics for FATKAT are finished and tested (and they work!), a prospective avenue for taking some Bayesian analysis into oilfield corrosion (a door I've been pushing on for a little while now) has opened up, and I had a very productive meeting with JCM seating on the postural front. So there's lots of good news this week.

Not every week can be like that, of course, but I'm hopeful that next week will see some important developments. Stay tuned!

No comments:

Post a Comment