Using Neural networks for Engineering problems

[babe]

I am contemplating using neural network software to help me find solutions to engineering problems. Does anyone have experience with neural networks and their application in engineering or R&D?<br>
<br>
What I am trying to find is the prediction of a response of a multivariable system outside the known parameter, without knowing the exact transfer function of the system. As an example, I would like to know what the weight of a wooden plank is when submerged for 5 years in alcohol of 50 C, having the data for a plank submerged for 5 years in alcohol of 20C and 1 year in alcohol of 50C. Stuff like that.<br>

 

[mandeep]

You can contact Alvaro at alvaro@rice.edu, he was my office mate at school and is knowledgeable on this subject.<br>
<br>
Mandeep

 

[SWUSER]

Hello!<br>
<br>
I have used Neural Nets (NN)to predict a response of problems multivariable systems in R&D. What you are trying to predict isn't too difficult. You will have to send a minimum and maximum for each variable that you will be using as your inputs. Coming up with those values will be require lots of experience in the field or you will have to rely on experience of others. If you have lots of data, more accurate your response will be. I used NN while i was in the navy to predict elastomeric material properties based on the fillers added to the the polymer and the processing conditions. Also, used NN to predict aircraft design parameters based on the data avaliable on similar aircrafts. The most interesting project was to predict the firing of the attitude adjustment jets on the space shuttle to bring it under control if it happen to be tumbling/rolling... etc. in other words out of control and the pilot or the commander were unable to bring it under control. We used inputs from the attitude sensors and taught the neural net on the fly to predict firing sequence of the jets to bring the shuttle under control.<br>
<br>
If you have any questions, please email me.<br>
<br>
Thanks.<br>
<br>
Kyle Ian Boeing

 

[bmayers]

Hi
I am using neural networks to approximate a groundwater numerical model for an optimization problem. Inputs are the flow rates, outputs are head and concentrations at a well location. My objective is to maximize Q while minimizing h and c for a coastal aquifer. Thinking of using a Genetic Algorithm for optimal solution. Any advice or reference to material?

Barry

 

[Rich2001]

I am using an adddon to excel for Neural networks to predict weld deposit's mechanical behavoir at low temperature.


If you want neural networks without writing the code try BrainCel with MS Excel. Braincell is fairly inexpensive, manual is decent.

Here is a link to the BrainCel site

http://www.palisade.com/html/braincel.html

To read an article, &quot;Using Monte Carlo Simulation for Pavement Cost Analysis&quot; go to:

http://www.tfhrc.gov/pubrds/nov00/lifecycle.htm

 

[TD2K]

Neural networks work really well when you have a) lots of data as previously pointed out to 'train' the neural net and b) the data 'surrounds' the conditions you want to analyze.

Neural networks do a lousy job, at the very best, in extrapolating data which is what I mean by my 'surround' comment. For example, if you had data on different types of wood planks submerged in alcohol at various temperatures up to 10 years, I wouldn't put much faith in anything the neural network came back if you asked for the performance at 20 years, that's just not how they work.

 

[arivolee]

hello i had done my UG project in NN and also doing PG project in same if u want details contact me to arivolee@operamail.com

 

[arivolee]

hi
it has been proved in many research works that the well trained NN can predict the parameters out the range for which it is trained.
so that u can use it and moreover for like this purposes the backpropagation neural network seems to earn good results.