Help with Rockwell Automation Ladder Logic for Max Height & Peak Detection

[*rodrigp*]

Hi everyone,

I'm working on a project that involves using a Keyence LR-X100 sensor in Studio 5000 V35 ladder logic to determine the object's orientation (right side up or upside down) and verifying the presence of all parts by counting peaks. Our object scan time ranges between 120 - 450ms, with expected max peaks fluctuating from 1 to 5.

We've encountered challenges with color consistency affecting our vision cameras, leading to inaccurate results. I'm conducting tests to explore if sensor-based detection can provide the reliability we seek in achieving these expected outcomes

Any insights, sample logic structures, or advice on utilizing Rockwell Automation software for this purpose would be highly appreciated! Thank you in advance for your assistance.

 

[drbitboy]

... challenges with color consistency ...leading to inaccurate results
... Any insights, ...

I recently applied a very robust algorithm developed for the New Horizons trajectory, which is more or less a straight line out of the solar system, to the Lucy trajectory, which orbits the Sun, to find the closest approach of a spacecraft to a planetary body. Guess what? It was not robust for Lucy.

The key to this will be understanding the character (quality, noise, consistency, possible range, timing, short- and long-term degradation, etc.) of the data. Maybe a filter would help, maybe it would not. Solving it in a PLC would not be my first choice, unless the data were very high quality. However, considering the image below I doubt that is the case.

Some folks on this forum will have experience with this kind of thing (french fry lengths come to mind), and may chime in. I doubt many of them solved it entirely in a PLC, which prefers a Pass/Fail input; I thought the selling point for Keyence was providing such a reliable Pass/Fail signal.

​

 

[drbitboy]

https://www.baeldung.com/cs/signal-peak-detection

 

[TheWaterboy]

French Fry length . . . I missed some fun not being in that industry.

 

[creativepaper]

If your images are repeatible and show a huge difference between lowest peak and highest valley, then why not count on the transition from peak to valley. Latch when you get above a threshold, then count when it dips below that threshold.
Put some hysteresis in to avoid any noise and you should be counting peaks.