Tuesday, March 20, 2012

Process Maintenance and Process Safety

The December issue of Hydrocarbon Processing features an editorial by Reliability Editor Heinz P. Bloch titled "Spare parts availability and the need for non-OEM options", which you can read here. In the editorial Bloch mention a somewhat dated article with facts about failures in chemical plants:
  • 25% of all failures are preventable but not prevented
  • 15% of all failures are predictable but not predicted
  • 20% of all failures are predicted but not acted upon to undertake repair
  • 25% of all failures are predicted and machines stopped to do repairs
  • 15% of all failures are neither preventable nor predictable
Since these data predates the turn of the mellinium, the question naturally is: Do they still hold? The short answer is yes! However, the definitions need to be updated, so here are the updated definitions:
  • 25% of all failures are preventable but not prevented because of an arbitrary decision that is simply not rooted in knowledge or experience
  • 15% of all failures are predictable but not predicted
  • 20% of all failures are predicted by not stopped to undertake repair e.g. because an experts request to shutdown was overruled by someone in authority
  • 25% of all failures are predicted and equipment is shut down for repair, but only with restorative maintenance efforts instead of proactive upgrades.
  • 1% of all failures are neither preventable nor predictable. Since human beings make the decision to build in earthquake zones and what building code to apply. Levies may or may not be built, maintained or not maintained. The 1% covers situations, where an event in a neighboring unit spreads to another unit, and result in a failure (the originating event is most like in one of the other categories!).
Now try to substitute either "process safety events" or "occupational safety events" for "failures" in the above descriptions. What would be the percentages in each category for those events? Do you think the percentage have changed over the last 20 years? Recall that during that period process safety managers have learned abbreviations like RMP, MOC, SIL, LOPA and many more.

Is more process safety always good?

Jalila Essaïdi discus the idea of making human skin tougher, so it e.g. can withstand the impact of a bullet. His work is part of the 2.6g 329m/s project, which concerns performance standard for bulletproof vests in collaboration with Forensic Gnenomics Consortium Netherlands.
On the surface it sounds good if soldiers could be better protected from the impact of bullets by toughening the skin, just like today they are wearing silk underwear as part of the their battle dress to protect against splinters from road side bombs. However the discussion on Jalila's blop site already shows, that the world is not that simple. One commenter remarks e.g. that the tougher skin also makes it more difficult for a doctor to get inside you to make life saving fixes.
I am just wondering if we can also get to much process safety at our chemical plants and if safety sometimes get in the way of   other things on the agenda, e.g. sustainability. Although on the surface process safety and sustainability should go hand in hand. But overall this discussion shows the importance of the MOC process now implemented in most process plants.

Thursday, March 15, 2012

Mainframes doing process control - again!

Last week an old friend and co-worker from my days in the Canadian petrochemical industry visited us here in Denmark for a couple of days. One of the days he and I drove to Kalundborg while the ladies were chatting and looking at Frederiksborg Castle.
Naturally giving our coming background in the oil industry the talk turned to process control computers. And my friend mentioned a new installation involving more than 50 servers, some properly vitualized. But certainly more than one box. My friend also told me, that one of his recent projects was life time extension of some mainframes for process control installed in the early 70's. That must have been either IBM 360 or IBM 370 systems. They properly cost a fortune at the time. But wait a minute! Think about the per year cost! These machines have been in constant use for about 30 years, and their useful life was being extended - properly by 5 or more years.
In light of this I said "I can't understand why one are not using modern editions of the mainframe like the z90 for  process control these days". The machine is powerful enough in the smallest possible edition to host hundreds of virtualized machines. And with the extender or sidecar there is even the possibility of have blades running Windows software if a particular application requires that.  My friend agreed, that this sounded like a good idea. I added that it would have the added benefit, that the hardware would have the muscle to run very complex real time simulations, which could be updated with real time process measurements. The possibilities appear limitless.
And I think the saving in cabling would be considerable. Of course their is a small problem. You will need engineers with a slightly different skill's, than those maintaining Intel based hardware. But you have the same education problem if your SCADA is change from Siemens to Honeywell to ABB.
What do you think?