PLM and industrial internet trend

Being blunt, the volume of data in even a big PLM installation is pretty small. A few 10′s of GB in the database, maybe a few 100 GB of CAD models…. When you start looking at sensor data collection and have to look at maybe TB’s a MONTH of new data, it’s clearly a very different environment. Personally, I don’t think PLM is the place to put this data or analyse it, BUT PLM is certainly the place to capture the results of the analysis and respond to the knowledge. If you consider a chemical plant with hundreds or thousands of sensored valves, pumps etc, a clear link is between the actual performance of the components and the design performance. That may be a mis-match in either a negative (failing to meet requirements) or positive (over-performing) direction – both of which could result in design change, plus the actual performance levels should be in the ‘MRO’ system (which may be part of the PLM toolset). The number-crunching involved in collecting and analysing that volume of data is way beyond a PLM installation, but there is clearly a need to interface the systems at some level to not lose the knowledge the data collection and analysis can uncover.

Cheers Oleg. I guess the point I was trying to get across is that because of the differences in scale and processing, we need to think about what level of connection – linking the raw data to the PLM record (e.g. holding sensor readings against an ‘as-maintained’ BoM for example) is not helpful, BUT… its this raw data that’s needed to (for example) calculate MTBF’s and hence derive maintenance schedules. I think we PLM ‘evangelists’ have a tendancy to want everything in the PLM system, but there is really no way to do this with genuine Big Data, so we need to think about what we need, where it comes from and, as you say, how best to link it. This tends to live very much in the MRO/As-Maintained world – not a traditional PLM strength – but very much where the requirements are borne out (or not!) and they are very definitely part of the PLM landscape.
I think there are two levels of linkage to be considered – the obvious ‘actual’ performance of an asset in the field – e.g. actual failure stats for serialised equipment. But also, the aggregated results for a CI – e.g. the overall performance of the model of pump in use in a specific design location – does it meet requirements. It starts to raise one very intersting question (of many) – do we need to serialise everything now we can collect data about it? I can just imagine the response to that suggestion in most engineering departments
I suppose a left field question here is whether you could run PLM on a Big Data technology – say Teradata, Netezza? After all, PLM doesn’t really require very fast response times…. and if you CAN co-locate the systems, that opens up some very interesting reporting and analytic opportunities which current PLM systems cannot offer.