Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!howland.reston.ans.net!torn!nott!emr1!pcain.mets.NRCan.gc.ca!pcain
From: pcain@cc2smtp.emr.ca (Peter Cain)
Subject: ANN's and Performance Prediction of Engineering Structures
X-Nntp-Posting-Host: pcain.mets.emr.ca
Message-ID: <pcain.93.000FC34E@cc2smtp.emr.ca>
Lines: 40
Sender: news@emr1.emr.ca
Organization: CBCRL
X-Newsreader: Trumpet for Windows [Version 1.0 Rev A]
Date: Mon, 11 Sep 1995 19:45:40 GMT

I am very new to the subject of ANN's, and only an engineer (mining) to boot. 
However, I have been contemplating the usefulness of ANN's in the prediction 
of underground opening stability based on "learning" from a large amount of 
data that I have compiled.

It would be very useful to predict the behaviour of underground openings in 
advance of construction so that additional support could be installed in areas 
where it was needed, and not throughout. In the application I have in mind, 
there is a considerable amount of information available on roof quality, 
installed support and reduction in height of the opening (it is the height 
reduction that causes problems with transportation and ventilation.

I would like to be able to use this data to predict the height reduction in 
planned roadways based on geological prognoses, various support regimes and 
other planned factors

I have obtained a couple of ANN packages through ftp (aiNet and WiNNN), both 
"crippleware"  as I believe the term is.These packages are rather expensive 
for speculative purchase, especially for my labs finances (government 
cutbacks, and all that).

Firstly, is it possible that an ANN would do any better than a regression 
analysis?

Secondly, does anyone out there have any suggestions as to how I might make 
the best of the packages that are available? (I am thinking of about 6 to 
10 input parameters) with possibly one to three output parameters for 
"prediction" purposes.

Thirdly, is there anyone out there who is working on any similar 
problems or has any references to this type of work.

Fourthly, are there any DOS or windows packages available that might be better 
than the ones I have, and lastly,

Thank you in advance for any help, and please forgive my ignorance on the 
subject.

Peter Cain
  
