Re: Automated Step Testing Software

Posted by Don Snowden on November 20, 19101 at 10:09:57:

In Reply to: Automated Step Testing Software posted by Nahum Lifshitz on November 13, 19101 at 11:19:50:

This is a very interesting area of discussion. I think you will find people in basically two camps: those who promote automated step testing and those who don't recommend it. But I also think that, hopefully, once engineers set down and review the fundamental requirements that the majority of the engineers in either camp won't be that much different.

Most experience practioners will only advocate automated step testing if you already have a reasonable guess what the model is, and therefor don't push the unit to a condition where the plant is violating a constraint. Thus I personally would only advocate automated testing if : a) already had a "decent" model of the plant and b) I was quite familiar with particularities of the respective plant operation.

You are exactly right in that an important result of open-loop testing and continuous engineer monitoring is to identify unmeasured disturbances AND, if possible MAKE them measured disturbances!

Personally speaking I think this topic is way oversold. 24-hour plant testing is really not all that cumbersome and we are getting better at it. In addition to learning more about the process, it also gives engineers a huge amount of credibility with operations which is very important during the remainder of the project. As far as reducing the manpower time, I don't think it will save much as the time one's spends trying to coordinate the automated testing they could already be well into the plant testing.

The one exception to my comment would be for controller revamps. If the engineer is already familiar with unit operations and there was a previous controller running then automated testing to refine the model is not a bad idea.

As far as using SmartTest, I haven't used it. Theoretically I don't see why there should be a problem. Correlating moves shouldn't be a problem. The only real issue will be that the data will be "biased" by the unmeasured disturbances, but they should have enough moves and "detrending" of the data to reduce the effects of the unmeasured disturbances.

Follow Ups:

• Re: Automated Step Testing Software Keith Landells 06:39:42 1/11/102 (0)

Post a Followup

Name    : 
E-Mail  : 
Subject : 

Comments:
: This is a very interesting area of discussion.  I think you will find people in basically two camps: those who promote automated step testing and those who don't recommend it.  But I also think that, hopefully, once engineers set down and review the fundamental requirements that the majority of the engineers in either camp won't be that much different.

: Most experience practioners will only advocate automated step testing if you already have a reasonable guess what the model is, and therefor don't push the unit to a condition where the plant is violating a constraint.  Thus I personally would only advocate automated testing if : a) already had a "decent" model of the plant and b) I was quite familiar with particularities of the respective plant operation.

: You are exactly right in that an important result of open-loop testing and continuous engineer monitoring is to identify unmeasured disturbances AND, if possible MAKE them measured disturbances!

: Personally speaking I think this topic is way oversold.  24-hour plant testing is really not all that cumbersome and we are getting better at it.  In addition to learning more about the process, it also gives engineers a huge amount of credibility with operations which is very important during the remainder of the project.  As far as reducing the manpower time, I don't think it will save much as the time one's spends trying to coordinate the automated testing they could already be well into the plant testing.

: The one exception to my comment would be for controller revamps.  If the engineer is already familiar with unit operations and there was a previous controller running then automated testing to refine the model is not a bad idea.

: As far as using SmartTest, I haven't used it.  Theoretically I don't see why there should be a problem.  Correlating moves shouldn't be a problem.  The only real issue will be that the data will be "biased" by the unmeasured disturbances, but they should have enough moves and "detrending" of the data to reduce the effects of the unmeasured disturbances.


Optional Link URL:  
Link Title:         
Optional Image URL: