Hi,

Does anyone know if there exists some form of Smith-predictor or predicting PI controller for S7-1500.

I am using TIA 13 and a 1511F CPU.

Hello ccduii;

I have found no S7 code (LAD or SCL) for a Smith Predictor function, unfortunately, on the Siemens site.

The only solution proposed is for PCS 7, programmed in CFC. But the documentation can probably help you design your own application in TIA Portal.

Smith Predictor for Control of Processes with Dead Times (PCS 7 v.7.1 blocks):

https://support.industry.siemens.com/cs/ww/en/view/37361207

There are a few discussions on the web descrtibing the results of the implementation  of a Smith Predictor algorithm on a S7-300/400 platform. Unfortunately no code seems easily available. But they still might give you enough hints to help develop your application on a S7-1500.

For example:

http://www.iaeng.org/publication/IMECS2010/IMECS2010_pp940-945.pdf

Hope this helps,

Daniel Chartier

Would increasing the sampling time of the PID controller produce a similar effect as using smith predictor?

I know it would not be as good but perhaps it can be good enough?

hello ccduii;

No, changing the sampling time of the PID controller will not help you approximate a Smith Predictor. Modifying sample times only impacts the quality of the mathematical calculations of the controller.

The relationship between the FB41 Cycle parameter and scantime (and its effect on performance) has been dioscussed many tiumes on the forum, for example here;

https://support.industry.siemens.com/tf/ww/en/posts/106780/

The theoretical impact of sampling time on a PID controller function is explained quite clearly here by Mr. Doug Cooper, from Contolguru.com:

http://controlguru.com/sample-time-impacts-controller-performance/

By the way, the Controlguru site has one of the best (to my knowledge) presentations of PID tuning methods, both for simple (self-regulating) processes and complex (non-regulating, presenting large delays) situations.  I suggest you look into their FOPDT (first-order plus deadtime) analyses and tuning examples (go up one level to the Table of Contents), this is the closest you will be able to approximate a predictor with PID parameters.

Hope this helps,

Daniel Chartier