I am interested to know whether "Statistical Process Control" (SPC)
has ever been applied to the "production" of data in an automatic data
logger. My question is, if so, what is the patent number or who
manufacturers the data logger? |
Request for Question Clarification by
ephraim-ga
on
27 May 2002 07:35 PDT
Can you give some specifics about what you mean by "automatic data
logger"? Are you referring to any device which continuously samples
some type of data without human intervention?
|
Request for Question Clarification by
ephraim-ga
on
27 May 2002 07:53 PDT
Another clarification -- can you explain what you mean by "applied to
the 'production' of data"? How is this different from using SPC to
control the acceptable bounds of the data so that you know when the
logger may need repairs or when the process being logged may need to
be reset?
|
Request for Question Clarification by
morris-ga
on
27 May 2002 14:58 PDT
I'm guessing that by "production" of data, you mean using SPC before
logging the data point, rather than downloading the data to a computer
and running companion SPC software. The latter case, free SPC software
with a data logger, I've seen on the web. I've also worked with some
of the upscale SCADA products, like Intellution, which give you full
SPC optionson the raw data stream, but I wouldn't call Intellutions a
data logger product.
|
Request for Question Clarification by
morris-ga
on
27 May 2002 15:58 PDT
Perhaps a better question would be, why would you want to run SPC on a
simple data logger? Memory is hardly an issue, and if the data is
simply being logged rather than used for real time control, what's the
advantage of live SPC? I actually found several products that
obviously integrate data logger capabilities and perform SPC, but they
are used in specific real time environments.
|
Clarification of Question by
mdalgleish-ga
on
28 May 2002 01:40 PDT
For ephraim-ga:
An automatic data logger records data with no human involvement.
Typically such a device might record temperature when placed in a
refigerated van for the entire journey, only being handled and checked
at the start and at the end of the trip.
It is not different to SPC on a production line in a factory. But a
key point is that if the data is quite truly varying in an erratice
manner, and the sensor is recording that, you will not know what the
"bounds of the data" should be. That is the point really of my
question, because I think I know a way to overcome that, and want to
know if its been done before.
For morris-ga:
Yes, I mean in the instrument itself.
The point of running it IN THE PRODUCT is that like in a factory, you
want to know when the "producing machine" has a fault, and as soon as
possible. Most control charts have that goal. So I want to know as
soon as possible whe the logger/sensor has a systematic fault, or
underlying change of process.
Thanks to all for your thoughts so far. This is very important to me
and I hope you have success.
|
mdalgleish-ga,
Thank you for using Google Answers.
Your clarification may have expanded the question from data
loggers into true process controllers, so I'll try to give examples of
both.
The Keithley Model 2700 Multimeter/Data Acquisition System is
a general purpose data logger that supports statistical operations in
real time.
"Virtual channel and onboard statistical analysis simplify real-time
application of mathematical functions to data. "
http://www.keithley.com:3080/PRODUCTS/modelNumber_tmpl.jsp?modelNumbers=2700
Getting a way from a simple data logger, the MARLIN (Machine
Reliability Inspection System) includes in the product description:
"Alarm conditions are also generated by violating onboard SPC
(Statistical Process Control) rules. These alarms automatically
provide a user with pre-programmed instructions to recommend and guide
corrective action."
http://www.skfcm.com/products/hardware/cm2243.pdf.
Another process analysis data logger from the Chemical industry
produced in Germany under the Knick brand. The English in the
description isn't ideal, but they appear to be using integrated SPC
functions for determining alarm conditions. I quote:
"In 1985, Knick presented the "Unical® 79" as the first system for
fully automated pH measurement. Even today, it is the only system that
permits unrestricted use in hazardous locations (Zone 1), can be
configured using an integrated SPC and can be accessed over standard
interfaces."
http://www.knick.de/1_sites/2_en/3_discourse/3_analyzers/
The same search also turned up the HPM MACO control system, for which
"You also get onboard SPC for any 10 parameters, with X-bar and
R-charts, histograms of current data and Cpk capability indexing with
USL and LSL alarms."
http://www.taylorsind.com/hpm/injection/universal/ucontrols2.html
Google Search Terms:
"onboard statistical" datalogger
"integrated SPC" data
built-in SPC
I actually helped a friend spec out a data logger for mass production
to meet a specific need in the PV industry. It hadn't occurred to me
that integrated SPC would be employed in production units. It is
certainly available with all PC based process controllers, such as
Intellution, which I've worked with quite a bit. Please post a request
for clarification if you require info about PC based systems.
I hope you find this information useful. |
Request for Answer Clarification by
mdalgleish-ga
on
28 May 2002 09:45 PDT
I am afraid my question has not been interpreted the way I meant. I am
sorry if I was not clear.
To be clear: I am concerned about the application of SPC to the data
logger itself, not the data being collected.
In other words, I wanted to hear about the use of SPC to indicate the
health of the measuring function in the sensor and data logger system.
The Knick device comes the closest, but it simply compares the state
of the electrode resistance with expected values to detect a known
failure mode. It does not seem to apply SPC to produce and error
warning or call for sensor replacement as far as I could see.
Can you address this point more precisely?
|
Clarification of Answer by
morris-ga
on
28 May 2002 10:20 PDT
Ah, the old "Who is watching the police" question. I'll take a look
around, but it is a very different question than what I was trying to
answer. I don't yet know how to retract an answer, but I'll e-mail
Google and suggest that they unlock the question again so other people
can take a shot at it. Given the breadth of the material I looked
through while attempting to answer the wrong question, I don't have
really high expectation that anything will turn up, but you never
know. Sorry if I got your hopes up.
morris-ga
|
Clarification of Answer by
morris-ga
on
28 May 2002 11:26 PDT
Question Clarification (since I posted an answer, the Question
Clarification option has vanished).
By shifting my search to plain "data acquisition" I ended up finding
more dataloggers with SPC, such as the following:
http://www.daytronic.com/products/4kspec.htm
Since that's not what you're looking for, I only mention it as a
search technique suggestion. As I read more and more and think back on
my own datalogger experience, I wonder if you and I aren't using SPC
in different ways, perhaps as a mathematician vs and engineer. From
the engineering standpoint, SPC is more than a statistical technique,
it implies actual Process Control is going on. Since I got my lowest
grade in Probability in grad school (EE), I may be oversimplifying the
issue, but if you perform process control by any means on your raw
data measurements BEFORE logging the data, it seems you are opening
yourself up to accidentally producing false data. SPC is a wonderful
tool for creating stable processes based on genuine data, but wouldn't
applying SPC to the actual data collection risk hiding spikes and
singularities that are part of the real world? Even if you only
trigger an alarm that the datalogger may be sick because the expected
data exceeds statistical expectations, I'm not sure what is being
gained. Maybe I'm still missing the point, and that's limiting my
ability to research the issue. Of course, lack of existence of
something on the Internet in the public domain isn't the same as a
patent search in any case.
|
Clarification of Answer by
morris-ga
on
29 May 2002 12:00 PDT
mdalgleish-ga
I'm afraid I don't have the ability to re-open the question for
you. Google informs me "Currently, answers are not removed at a
researcher's request." The only way you can re-open the question is to
reject my answer. If you still want other Google Researchers to work
on it, I'd suggest reposting the question with some different wording.
|
