BookmarkSubscribeSubscribe to RSS Feed

What is meant by JMP Alert - Profiler will be faster by decreasing Default N Levels

Highlighted
dlifke

Community Trekker

Joined:

Aug 6, 2013

I am creating a fairly messy reliability block diagram in JMP Pro, so I'm using Library Items to keep my top level Design Diagram cleaner.  As I add parallel redundancies within my Library Item, which is then used multiple times in parallel redundancies in my Design Diagram, things really slow down and eventually stop. Prior to JMP stopping, I received the JMP Alert "Profiler will be faster by decreasing Default N Levels." I could not figure out what that meant or how to do it.  Any suggestions?

 

Thanks!  Don 

1 ACCEPTED SOLUTION

Accepted Solutions
peng_liu

Staff

Joined:

Jul 3, 2014

Solution

These are what I can think of:

1) Under the "Start" block, there is a check box "Run". Uncheck it, so you can continue editing the diagram, but the calculation will not be carried out on the fly. Check the box, after finishing editing the diagram.

2) If the parallel redundant components are identical, use the built-in "Parallel" block. If the parallel redundant components are sub-systems, there is no way to put sub-system in "Parallel" block for now. A workaround is to use the "Nonparametric" distribution in "Parallel". First come up with a data set, whose nonparametric estimate of the distribution is close to the distribution of the subsystem. Then feed that data to Nonparametric distribution in "Parallel" block. Coming up with the data set is not a trivial task.

3) Use "Repairable Systems Simulation" platform. The "Point Estimation of System Availability Profiler" is the reliability function, if no maintenance is involved. This approach is slower for small systems, but more feasible for large systems. However this platform focus on different metrics, and does not provide the same set of graphs as those in "Reliability Block Diagram" platform. Also this platform does not support sub-system yet.

7 REPLIES
Craige_Hales

Staff

Joined:

Mar 21, 2013

outside of my area...but I found this example in the scripting index

 

Capture.PNGScripting index

Craige
Jeff_Perkinson

Community Manager

Joined:

Jun 23, 2011

JMPScreenSnapz259.png

The curves in the Prediction Profiler are formed by evaluating the the prediction formula a number of times for each factor across the range of the factor while holding the other factors constant at their current values. These points are then connected to form the curve. The number of times the formula is evaluated for each factor is controlled by the Default N Levels option under the red triangle.

 

JMPScreenSnapz260.png

 

From the help page:

 

Default N Levels
Enables you to set the default number of levels for each continuous factor. This option is useful when the Prediction Profiler is especially large. When calculating the traces for the first time, JMP measures how long it takes. If this time is greater than three seconds, you are alerted that decreasing the Default N Levels speeds up the calculations. 

 

-Jeff
dlifke

Community Trekker

Joined:

Aug 6, 2013

Thanks for the quick reply Jeff.  I am in the Reliability Block Diagram in JMP Pro, so there is not a Prediction Profiler.  There is a Distribution, whichI guess is essentially a profiler, and a Reliability plot, which also is probably a prediction profiler.  Under the red hotspot for Distribution, then Reset Factor Grid, there are options for Current Value, Minimum Setting, Maximum Setting, Number of Plotted Points, and checkboxes for Show and Lock Factor Setting.  I changed the Number of Plotted Points from the default of 41 down to 5, but that didn't seem to stop JMP from Not Responding.  Each time I changed something in the block diagram JMP seemed to want to recalculate, which is fine. However, each time I added a new block, it got slower and slower. Eventually it just stopped.  I let it run overnight and it did seem to finish whatever it was doing. However, when I made another change, it went right back to Not Responding. Now the file takes a long time to open, and JMP eventually shuts down. I'm guessing I need to reduce Number of Plotted Points right away for all Distribution and Reliability plots, before adding any additional blocks. Do you think that might resolve the issue?  Again, I greatly appreciate the help!

Jeff_Perkinson

Community Manager

Joined:

Jun 23, 2011

Yes, the Distribution outline node in the Reliability Block Diagram is actually a Prediction Profiler.

 

It sounds like you do have a really big, complicated diagram that JMP is having difficulty with. Tech Support is best positioned to help figure out if there's anything that can be done to speed it up. 

-Jeff
peng_liu

Staff

Joined:

Jul 3, 2014

Solution

These are what I can think of:

1) Under the "Start" block, there is a check box "Run". Uncheck it, so you can continue editing the diagram, but the calculation will not be carried out on the fly. Check the box, after finishing editing the diagram.

2) If the parallel redundant components are identical, use the built-in "Parallel" block. If the parallel redundant components are sub-systems, there is no way to put sub-system in "Parallel" block for now. A workaround is to use the "Nonparametric" distribution in "Parallel". First come up with a data set, whose nonparametric estimate of the distribution is close to the distribution of the subsystem. Then feed that data to Nonparametric distribution in "Parallel" block. Coming up with the data set is not a trivial task.

3) Use "Repairable Systems Simulation" platform. The "Point Estimation of System Availability Profiler" is the reliability function, if no maintenance is involved. This approach is slower for small systems, but more feasible for large systems. However this platform focus on different metrics, and does not provide the same set of graphs as those in "Reliability Block Diagram" platform. Also this platform does not support sub-system yet.

dlifke

Community Trekker

Joined:

Aug 6, 2013

Block_Diagram_example.pngSampel Block Diagram

 

Thank you for all the helpful replies! I greatly appreciate your time.

 

I tried just unselecting Show Distribution Profiler to speed things up, but calculations were still apparently taking place behind the scenes.  Unchecking the Run button did the trick, although when I eventually recheck it the calculations still take quite a while to run, even with N = 3. After the initial calculations, slowly increasing N to 7 and then to 15 resulted in much faster calculations. I did get rid of the Library object in my diagram (another complex parallel series system) and replaced it with K out of N blocks as approximations (not perfect, but still useful). That seems to prevent JMP from crashing.  I attached a similar diagram showing the complexity, just for reference. It may be the networking-style of redundancies that are taking a while to calculate. I certainly could never calculate them by hand. Thank goodness for JMP Pro!

peng_liu

Staff

Joined:

Jul 3, 2014

Glad that helps. And thanks for the example. The example will help for future improvement.