Analytical Instrument Reliability

In an application where we have an off-line analytical instrument looking at a fluid stream, we are cleaning the sensor every week
(deposits on the sensor degrade its performance). Even with that, the instrument is failing to respond to a check signal 10 % of the time. Are there any standards or guides (ISA, IEEE, etc.) that quantify whether 10 % is acceptable or not acceptable? My experience tells me this is very poor performance, however, convincing the money men to look at new technology will be difficult. TIA
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@srs.gov says...

Are you leaving the sensor in the fluid stream when it is NOT in use? What is the nature of the deposits? Is the sensor designed to measure clean fluids, and you are submitting it to a slurry?
If you want to convince a "money man" of anything, show him THE MONEY. If you are spending 1 hour per week cleaning the instrument, then: (50 x hourly rate) ~= annual maintenance cost.
Let's say, for the sake of argument, that your cost to the dept. (salary, bennies, overhead) works out to $100/hour. That's $5,000 per year. What does the new technology cost? If it's less than $5K, you'll have an easier case than if it's $50K...
--Gene
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Gene S. Berkowitz wrote:

A true money man will ask what else he might use the time for. "Noticing an impending disaster in time to avert it" will leave many bean counters unconvinced. Fearless Leader stated early in his tenure that we couldn't afford to seriously address greenhouse gases. He and the rest of the bean counters are beginning to learn that we can't afford not to.
Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Sensor has to remain in the fluid stream for real-time monitoring (safety application)

The sensor is designed to detect small (micron range, can't be more specific) particulate in the fluid stream. The deposits are the particulate accumulating on the face of the sensor

We have already quantified what it is costing us in $ to maintain these instruments. The new technology would have to be developed (R&D, of a sort) with unknown $. This risk of missing a signal is far higher ($, as well as negative exposure, which we can't quantify in $) which is why I wanted to go after a reliability argument

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@srs.gov says...

Perhaps English isn't really my native language, but that certainly strikes me as NOT "off-line"...

Can the sensor operate correctly in turbulent flow? It might reduce the amount of accumulation, although it sounds like there's a material incompatibility if the particles intended to be measured bind easily to the sensor face.

What does the OEM say about all this? Are they still in business? Was this a one-off analyzer?
--Gene
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Maybe I should have described it as "not in-line" (this is a 24-inch diameter main line). A sample stream is continuously pulled out of the main line

The material compatibility issue is what I want to investigate with R&D $ from the money men

This was a one-off design developed in-house ~ 40 years ago. Believe it or not, extensive efforts have failed to uncover a single manufacturer anywhere making a commercial version, perhaps due to the fact that nobody else uses this instrumentation in this application

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@srs.gov wrote:
...

...
Can you sample the side stream intermittently? It's a long time ago, now, but a particular measurement -- pH or BOD -- was made manually every hour around the clock at a sewage treatment plant. It was claimed that the measurement couldn't be automated because continuous immersion wouldn't work. The sensor had to be cleaned between measurements. I devised a system that put the sensor in a pipe that extended into the treatment tank. Between measurements, a small flow of recirculated final effluent bathed and rinsed the sensor. To take a reading, a three-way valve put the sensor in contact with the tank's contents. The system remained in use for several trouble-free years and was replaced during an overhaul of the entire instrumentation package.
Jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@srs.gov wrote:
...

What does "off-line" mean to you? How often do you record a measurement? How often do you need to record a measurement? What is the settling time of your sensor? Would the stream being analyzed be contaminated by a flushing solvent? Depending on the answers, I may have a solution for you.
jerry
--
Engineering is the art of making what you want from things you can get.

  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Maybe I should have said "not in-line" (main line is 24-inch dia.). A sample stream is continuously pulled off the main line. Recording is continuous in real-time and must remain so (except during maintenance). Current maintenance includes a system flush with a light acid solution. Depending on the solvent, contamination may not be a problem. This is getting OT to my original question so feel free to take this to e-mail.


Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

The answer is 'it depends' on many factors. In my experience, that sort of availability and confidence is pretty marginal. Is the check process automated and can the failure be flagged and the erroneous result blocked? If so, is more frequent validation an option? What is the impact of an erroneous result getting through?
If the intent is to persist with this thing then there must be a reason. If it's 'important' then that needs to be properly quantified and promoted (latter of these more important IMO). If it's not, why spend time on it at all?
Marginally useful analytical measurements have a strong tendency to drop off the perch as sponsors drift away and people get tired of keeping them running for no clear reason.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Bruce Varley wrote:

If an unremarked bad reading causes harm outside the plant walls, what is the company's liability? If it is learned that management deemed bad readings acceptable, how does that affect liability? The money man may know, but the legal department might be a better source.
Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Jerry Avins wrote:

Jerry, are you saying that the bean counters can see dollars, but can't always see sense?
--

Tim Wescott
Wescott Design Services
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Tim Wescott wrote:
...

Sort of. When a new bean counter at RCA Labs had us documenting our time to the nearest quarter-hour throughout the day, I began charging fifteen minutes a day to overhead. After a few weeks, that was questioned and I explained it was time I used for filling out time cards. The order was withdrawn. The Big Cheese commented to me a few days later, "I knew I could count on you." The bean counter was replaced after a few months. I was very lucky, working where I did.
Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

The check process is manual and failures are flagged, which is where the 10% value comes from. More frequent validation means more manpower ($ of course). The impact of an erroneous result escaping detection is very high ($, as well as negative exposure), but hard to quantify in just $, which is why I was hoping to find some published reliability numbers.

The application is a safety system. The risk of missing a true signal is difficult to quantify in just $.

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@srs.gov wrote:
...

In my line of work, a computer crash could mean blood (with luck, only broken glass) on the factory floor. That too is difficulty to quantify, but its cost is high. Fortunately, I never had the opportunity to gather data on how high, but once I nearly had to resign in order to remove the money constraints that might have created that opportunity.
Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.