Loss-In_Weight Calculation in ControLogix


Thread Starter

Frank Jacobs

I am trying to find a good way to calculate the loss-in-weight from a powder feed system. Currently i am using an A-B ControLogix PLC with
a RIO link to a weigh scale interface. The method I am using uses a set time sample with a filtering routine. The calculation currently
overreacts to changes in weight but if my filtering algorithm is increased the process reacts too slowly. Any help would be appreciated.

Anthony Kerstens

Does the scale have it's own built-in filter that could allow you to remove your's?

Anthony Kerstens P.Eng.

Alan Rimmington

Sounds like the instrument resolution is too low or your method of predicting the weight change is flawed.
Try using an algorithm which looks for perturbations which are impossible and then freezes the control until the perturbation goes away. In other words, use a rate of change filter to place the loop into Manual Mode.


George (Jim) Hebbard

I, too, am interested in L.I.W. control algorithms. I know nothing about ControlLogix, but have faced the same problems using Visual Basic and a Metler-Toledo RS-232C scale interface. Because the scale (balance) is only 1 decimal place, and the readings are taken every second at less than a gram per second rate, there is a >lot< of noise. We have overcome it well enough by using 10 second rolling averages for all controls calculations.

Anyone who knows of any free or public domain information on Loss In Weight applications, please chime in and give us some guidance or references.


As I tried to imply in my last note, the important thing is NOT to filter the noise in to the signal but REJECT the signal when it has an adverse amount of noise. Most filtering algorithms are smoothing algorithms which means they allow some remnant of the noise to be added into the controlled variable. Loss in Weight systems are typically hopper based and prone to vibration from a variety of sources. The simplest method is just to monitor the first derivative of the PV and freeze the input and PID calc when the first derivative is too large.


I work for the field service department of a loss-in-weight feeder manufacturer and I receive many requests for information on the LIW
algorithm of course, I can't give the one that we use. My company has spent more then 20 years developing digital LIW feeders and our
controls can interface with many of the popular PLCs and HMIs. Due to the complexity of loss-in-weight feeding, I believe that it is simpler
to interface with a dedicated LIW controller then to attempt to construct a PLC algorithm.

However, your problem of an over reacting scale signal can be caused by a poor mechanical installation and not the sensitivity of the PLC algorithm. I see this problem all the time. A LIW scale has to be free of any restrictions. A simple test is to verify the scale repeatability. This is done by noting the scale position, pushing down on the scale and releasing, and noting the scale position again. If the two reading do not agree (at least very closely); there is some type of mechanical restriction. Remember to repeat this test a few times and do it for both directions (lift up and release). If it doesn't repeat, start disconnecting any type of connections that travel from the frame to the scale. Usually, this will solve this type of problem.
If you have any questions, please contact me.

I've never been able to find free/PD info on how different manufacturers implement loss weight algorithms, and suspect all the juicy, non-trivial items tend to remain proprietary.

There's a couple of overview articles on the Powder and Bulk website,"http://www.powderandbulk.com/pb_services/ask_joe_archive/loss_in_weight_gain_in_weight_feeders.htm":http://www.powderandbulk.com/pb_services/ask_joe_archive/loss_in_weight_gain_in_weight_feeders.htm ,


but they don't go into much detail.
Joe O is right. PLC does not really have sophisticated enough math functions to attack this problem. To create the function you need would require a lot of information on the scales response. Remember impulse response from your
engineering classes? If you have this info than you can apply the sophisticated mathematical techniques you learned in your control loop design
theory classes ;) Short of that monitor the first or second derivative and simply set a limit that removes the oscillations from the input of your control loop. This approach however will require ---well---guessing.
I suggest you to use the LOSS IN WEIGHR WEIGHERS and connecte the signal in 4-20 mA to PLC. Use Buhler Loss in weight weighers for Powders. Should you need information contact me with details on e mail: [email protected]