Quality, reliability and data...
Aarnout Brombacher · Quality and Reliability Engineering International · 2018
When I was working on my PhD, in the late 1980's, both my supervisors insisted that I would carefully monitor and describe my experiments and also that I should carefully keep the products and devices that I used in these experiments so that, if in the future anybody would be interested to re-do my experiments that he/she could do that. Recently, when cleaning up the attick in my house I came accross these devices so if there is anyone interested to do reliability tests with 30 year old power transistors: just let me know! This sounds perhaps like a silly anecdote but it touches upon a major problem that many contemporary researchers are faced with. If necessary I could re-do my original experiments because the devices and relating products are still available. If I would like to add these new results to my original thesis I would be faced with a more serious problem. Altough I also still have my old floppy-disks these disks are absolutely useless. Neither the hardware (IBM-XT) nor the software (Windows 3.11/Ventura) that I used in those days is still available. The problem that I would like to point out is that, on one hand, there is in the academic world a common understanding that any future researcher should be able to re-do (or further develop) experiments by earlier researchers. This requires thorough documentation of the research projects performed as is common with research papers; also with the papers published in this journal. The problem is that research projects increasingly become dependent on ICT; in the measurement equipment used, in the analysis and/or simulations performed and in the storage of the resulting data. For the more mathematical papers this is probably not too much of a problem; mathematical descriptions have proven to be robust for a very long time. For the more engineering oriented papers this problem can become considerable. Quite often we see papers with the (in)famous sentence “given a dataset X” where it is absolutely unclear when, how and with what this dataset was obtained. This makes it very difficult to confirm the integrity of the dataset. Developing follow-up experiments is often even impossible. In the Netherlands there has even been a case of a professor at a reputable university who for many years “invented his own data” which went unnoticed for at least a decade. Therefore I would strongly advocate that people, especially in engineering oriented research, adopt the so-called FAIR principles; for all papers the underlying data should be Findable, Accessible, Interoperable, and Reusable. If I look to the currently submitted papers we still have a way to go…