NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/comp/ai/neuraln/3199

Internet Message Format | 1992-08-14 | 5.6 KB

Path: sparky!uunet!dtix!mimsy!afterlife!hcbarth
From: [e-mail] (Bart Bartholomew)
Newsgroups: comp.ai.neural-nets
Subject: Re: Advice re suitabilty of ANN's for this application
Message-ID: <[e-mail]>
Date: 15 Aug 92 05:17:39 GMT
References: <[e-mail]>
Organization: The Great Beyond
Lines: 109

In article <[e-mail]> [e-mail] writes:
>
>
>We are trying to determine if the following problem is one which is suitable
>for addressing with a neural network. If so, what sort of network would be
>"best" to use? Any advice re: software to use, implementation and testing
>of the system, relevant literature etc would be greatly appreciated.
>
>The problem may be stated as follows:
>
>In a nuclear power reactor, a wide variety of chemical processes may occur in
>the water used both in the boiler and the steam side of the heat transport
>system. Important chemical parameters to be monitorred include the pH,
>conductivity, dissolved hydrazine, dissolved oxygen, and particulate
>concentrations. Often, transient conditions occur, during changes in reactor
>power, which lead to sudden changes in the water chemistry.
>These chemical processes can lead corrosion and the formation
>of "crud", leading to boiler fouling. This in turn leads to lost revenue
>due to derating, shutdown, cleaning and replacement of components.
>
>Monitorring of the chemistry has been in progress for a number of years,
>primarily through the acquisition and analysis of grab samples, although more
>online monitoring is being included as time progresses. The reactor
>operators are required to monitor the chemistry in the heat transport
>systems, to attempt to identify when an undesirable event is likely to
>occur, and to take preventive measures to minimize damage and lost revenue.
>The process of trying to assimilate the large amounts of data, and to
>qualitatively determine whether this will lead to an undesirable event, is
>of course quite difficult.
>
>We have as data (for a potential training set) the chemistry related data
>from before, after and during a number of "events". Also included in the
>data sets is the efforts taken by the operators to deal with the event,
>and info on what caused the event to happen. Two approaches have
>been proposed to make use of this data. One is to develop a mechanistic
>model based on the chemical reaction kinetics of the species and surfaces
>present in the system. This model could then be used to determine the
>conditions under which fouling is likely to occur. The second approach would
>be to use a non-mechanistic approach, such as an Artificial Neural Network,
>which would be trained, using the data described previously, in such a way
>that the data for the "current" measurements, and perhaps some information
>from "past" measurements, could be used as input to the network, and the
>network would provide as output:
>
>    a) Problem  / No Problem    (minimal information needed)
>
>         This would be what we would consider successful, but not
>         very satisfactory solution to the problem
>
>    b) Problem Identification   (ie there is a problem due to the combined
>                     effects of high dissolved oxygen and
>                             low pH {this type of information to
>                         have been obtained from the training
>                     sets}     )
>
>    This would be both successful and satisfactory if achieved.
>
>    c) Problem Identification and/or Solution Proposal
>
>        (ie. the network identifies the problem and proposes
>             a method, based on past experience, for resolution
>             of the problem)
>
>    This would be the ideal solution to the problem.
>
>The main desire in this program is provide an automated method for dealing
>with the data to minimize the need of the reactor operators to concern
>themselves with the chemistry of the heat transport loop, and focus even
>more of there attention of reactor operating conditions and safety (which is
>already there primary responsibility, they would prefer not to have to worry
>to much about the chemistry end of the heat transport)
>
>An expert system approach has been attempted by others for solving this
>problem, but it apparently didn't meet with much success. Hard to get much
>info about it, as it was done under a proprietary research contract to an
>electrical utility.
>
>ANY ADVICE/SUGGESTIONS/INFORMATION re the feasibility of this project
>would be greatly appreciated.
>
>                Thanks in advance
>                    Phil Harris

    Even as I type this, without having read the remaining
articles, I just *know* that Prof Armstrong is similarly entering
a reply.  The thing is, I think he will be right.  This type of
problem looks like a natural for ALNs.  The CoDomain feature
of the ALN is the best approach I know of for the answer to
the questions
    1: Is there a problem?
    2: If so, what is it?
The answer to 2 would presumably lead easily to the required
corrective action.
    I'm also sure that Prof Armstrong would emphasize
that the network output should be construed *only* as advisory,
perhaps spotting a problem earlier than otherwise, perhaps
confirming a decision that the operators would make anyway.
    BackProp could probably answer #1, but I don't see how
to get it to answer #2, unless you have a series of small nets
feeding a final decision, and then you see which of the smaller
nets set things off.
    Prof Armstrong, it looks like this might be the problem
you have dreamed of.
    Bart
--
"It's not the thing you fling, the fling's the thing." - Chris Stevens
 If there's one thing I just can't stand, it's intolerance.
*No One* is responsible for my views, I'm a committee. Please do not
infer that which I do not imply.  [e-mail]