Re: [foxboro] Measuring Overall System Reliability

I unfortunately don't have the be-all, end-all measure of reliability, but 
I can say that anything that causes loss of control and/or view of the 
process gets our attention.  I/A seems to have a higher frequency of 
system faults than the other DCS I have spent a lot of time with, but 
redundancy in the system has prevented lost control/lost view in most 
cases (but not all...).  Things like the CP60 fieldbus errors (which we 
have as well), NMI errors that cause one-side CP reboots, etc.  are 
usually watched carefully, and offending hardware is replaced if the 
errors get too frequent.
This is an interesting idea, however.  A script or three could generate 
histograms and statistics on hardware by letterbug or something by 
grokking the system monitor reports.

I don't subscribe to the 100% uptime notion for any system, and I/A is 
among them.  Day 0 upgrades alone mean that my plant would still be 
running version 4.0 if it didn't have turnarounds every 1-2 years.  But 
then again that 100% thing could be management's idea, and not yours :)

Corey Clingo
BASF Corp.






tom.vandewater@xxxxxxxxxxxxxx
Sent by: foxboro-bounce@xxxxxxxxxxxxx
06/14/2004 12:47 PM
Please respond to foxboro

              To:  foxboro 
              cc: 
         Subject:       [foxboro] Measuring Overall System Reliability






Hi List,
I am trying to develop a measure of overall control system
reliability and wonder if anyone out there already has developed a metric
for this purpose. Our twenty node, carrierband connected system is as
redundant/fault tolerant as we can make it with the current Foxboro
offerings with the exception of redundant I/O.  We use dual carrierband,
dual nodebus and dual fieldbus functionality as well as dual power sources
with UPS backup and are expected to provide 100% on-line time for our
24x7x365 operation.  We also climate control almost every nest room where
the FT controllers and non-FT I/O are located.  We have experienced pretty
good reliability with the exception of individual FBM failures that 
control
critical process variables or more recently, fieldbus communication 
failures
related to Foxboro's new FCM, DCM, FBI10E's.  Foxboro continues to search
for the solution to that problem with new EEPROM updates being issued 
every
three or four months and we have only experienced one loss of 
communication
that resulted in a process downing event, but volumes of intermittent
fieldbus communication errors.
What I am looking for is a measurement that we can continue to track
that will indicate when system reliability starts to deteriorate.  Since
2/3rds of our installed base still uses legacy FBM's that will need to be
replaced sometime in the future, I have started to focus on them.  We have
recorded failure data on them since our initial 12 node installation in
1989.  We saw an initial infant mortality of those modules in 1989-91 and
then the rate dropped off and continues to track along the bottom of the
bathtub curve at an acceptable electronic module failure rate of less than
.5%/year.  We have seen an increasing number of those modules that would
run, but failed to successfully burn new EEPROM updates during software
upgrades.  We know that the old FBM's will eventually start failing at a
higher rate and climb up the other side of the bathtub curve but we don't
know when.  If we had some good prediction of that increase in failure 
rate,
it could be used to justify to management the significant budgeting needs
that will be required to replace all of the legacy FBM's.
When Invensys acquired TRICONEX I had hope that Foxboro and TRICONEX
would pool their resources and come up with a Triple-Modular-Redundant,
(TMR), I/O module that could communicate directly to Foxboro or Triconex
controllers but I have seen no visible activity in that area.
If anyone has any ideas about an effective measure of system
reliability please contact me off-list or share the info with the rest of
the users on the list.

Thanks in advance,
Tom VandeWater
Control Systems Developer/Analyst
Dow Corning Corp.
Carrollton, KY   USA






 
 
_______________________________________________________________________
This mailing list is neither sponsored nor endorsed by Invensys Process
Systems (formerly The Foxboro Company). Use the info you obtain here at
your own risks. Read http://www.thecassandraproject.org/disclaimer.html
 
foxboro mailing list:             http://www.freelists.org/list/foxboro
to subscribe:         mailto:foxboro-request@xxxxxxxxxxxxx?subject=join
to unsubscribe:      mailto:foxboro-request@xxxxxxxxxxxxx?subject=leave

Other related posts: