Why have system models????

I' m confused, I spend all this time at school modeling systems. I don't care for it model much, Ok it's nice to have a model of a system
but how is it going to help me in everyday industry.
I have two questions:
Do I need a model for my system to ensure that it will be stable under all conditions?
When I tune (Z-N method) a PID controller does that guarantee system stability?
Thanks!
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Cosmo wrote:

That's an irrelevant question. You can't guarantee that your system will be stable under _all_ conditions.
You can come close, though, and for most of that I use a combination of a system model that's been verified through measurements. For the work that I do these measurements are swept-sine. For some other systems (anything with lots of friction and/or backlash) bump testing with ARMA fitting may be more appropriate. Some things I do are so nonlinear, and the control goals are so far from what could be achieved, that I get acceptable results just designing from first principals.

No. This is a pretty basic fact of life with Z-N, that's out there to find with some pretty quick web searches.
--

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

How long is that? A class or two?

1. Models can help you avoid big mistakes. 2. Models can show what items are going to be the limiting factors in a design. 3. Models can help one optimize designs.
Most often I, or my customers, use models for auto tuning motion controllers. These are derived by correlating the control signal to the feed back. The motion controller setup software hides to computations but not the resulting model. This is all after the systems is designed and hopefully it was done right. The person installing the motion controller appreciates the time saved by not having to tune the system himself.
The big money saver is modeling the system before it is built. I don't like to use models to prove a design will work. There are too many un-modeled 'features' that can be more significant than originally thought. However, just having a model that shows a design will not work is easy and still very useful if you understand why the model didn't work..
I have seen people or companies lose contracts because they could not prove their design would work and the other could. In this case modeling is an important part of marketing and sales. What company do you want to work for?
Your other questions aren't as important.

No, you can always be very conservative. There are many industries where designing is done on rules of thumb. Normally these rules apply to systems that have been done over and over and over again so you can be pretty sure they will work. However, in this competitive world the limits are always be stretched and the rules of thumb don't always apply. So, do you want to be on the cutting edge or do the same thing the same way it has always been done? BORING.
Peter Nachtwey Some one who takes modeling and simulation seriously.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Ya, a class or two or three is about right. I plan on working in the nuclear industry, I've worked there as a trades man and wish to continue as an engineer when I graduate. I don't find my job boring, even if it is old technology we use, I enjoy ensuring the plant runs as it was designed to. I'll be in the maintenance department when I start work, and i'm trying to see where a model may help me at my job. I guess I see it in circuits with p-spice and EWB but i have yet to see a model for a temperature loop that considers all the temperature loses, the inputs and outputs to the system. HOw complex would these models be, can they be simulated.
Thanks for your
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Cosmo wrote:

Then I assume you are graduating with a degree in nuclear engineering.
If you are working at a nuclear plant then you will working in an operational capacity and not as a designer. Then you don't need to know modeling, I didn't, and can concentrate on procedures and maintenance. Believe me, these are just as important because you don't want to screw up anywhere on anything but you probably know that. I hated operational nuclear power and prefer real engineering, modeling and simulation. Different strokes for different folks.

The temperature models probably aren't that complicated. If you are just interested in temperature simulations like heat exchangers then see http://www.controlguru.com
I would bet the simulations of the free neutron density within a core of the reactor and the formula for sustain nuclear reactions would be a more complicated yet someone managed to get it right back in the 50s.
I would also bet that before you are qualified to operated a plant that you will know the information you need to simulate many of the systems. When I was in operational nuclear power there was a lot of emphasis on knowing or having a gut feel for what the plant is doing or will be doing so mistakes like TMI never occurred.
Peter Nachtwey
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

IMO, modelling when you don't have a sound hands-on understanding of the real thing you're working with is potentially very misleading. Mathematical models can easily appear OK, and be constructed in accordance with the fundamental principles that *should* apply, but behave very differently to the real world. It's only when you have the experience to pick the possibilities of mismatch that you can gain some confidence in what you're modelling. If you're new to the business, my earnest advice is to work with the real world for as long and as intensively as you can. Good luck.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Bruce Varley wrote:

And even when you're old and gray, _check_ your models against reality when you're done.
--

Tim Wescott
Wescott Design Services
  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.