SiePortal

The integrated platform for your product selection, buying and support workflow - bringing together Industry Mall and Online Support.

Products & Services
- Menu temporary not available
- Menue temporary not available
- {{item.title}}
Support Support
{{(apiData.Support | filter:{'rank':1}:true)[0].title}} Support
- Menu temporary not available
- Menue temporary not available
- {{item.title}}
MySiePortal MySiePortal
{{(apiData.MySiePortal | filter:{'rank':1}:true)[0].title}} MySiePortal
- Menu temporary not available
- Menue temporary not available
- {{item.title}}
Cart ({{apiData.CartCount}})

Navigation

General Question regarding Neural Networks / Fuzzy Control

8/6/2009 10:23 AM	Rate (0)
DesignTech Experienced Member Joined: 9/6/2008 Last visit: 6/9/2025 Posts: 178 Rating: (2)	Hi All We have a customer who has WinCC v7.0 with Redundancy/Clients etc... communicating to two S7-400 PLC's. He has an application in one section of his plant that involves blending/reacting of three products via flow control. This blend goes to an absorber which in turn is under level control controlling the inlet flows. This blended product moves progressively through another two absorbers which are under level control. In the final absorber the product is 'blended' with a gas which ultimately turns it into the finish product but also removes a heavy metal from it to make it safe for storage and further processing. If the product is high in heavy metals it remains in circulation until the gas is absorbed enough. Like most production goals the aim is to do all of this blending, absorbing and 'cleaning' as best possible with no circulation to give maximum throughput. This smacks of a neural network / fuzzy application. Does anyone have advice, where do we start, what packages do we need, fuzzy control or neural networks (is there a difference) Any advice would be great appreciated. Regards Philip

Suggestion To thank Quote Answer

8/6/2009 7:58 PM	Rate (1)
Wüstenräuber Gold Member Joined: 10/3/2008 Last visit: 9/8/2025 Posts: 996 Rating: (129)	Hi, FuzzyControl++ is recommended when human knowledge (rules) are used as a basis, and then transformed into a set of fuzzy rules. NeuroSystems are used to learn from existing data sets. Neural networks are able to model (high dimensional) technical systems and allow to control processes without the need for any detailed knowledge or complex mathematical description. As a consequence FuzzyControl++ and NeuroSystems can be combined very well. More detailed information can be found here: http://www.siemens.com/neurosystems/ http://www.siemens.com/fuzzycontrol/ For more details please contact mailto: it4.industry@siemens.com Wüstenräuber
	Last edited by: Wuestenraeuber at: 8/6/2009 7:59 PM
Suggestion To thank Quote Answer

11/17/2011 3:50 PM	Rate (0)
Dr. Neural Network Posts: 1 Rating: (0)	I would suggest the so called "Bounded Derivative Neural Networks". They take into account the relations which can be expressed as monotonicity rules. We had a project @ CT T Ru with these networks. they performed very well. I can give a paper on it, feel free to contact me.
Dr. Neural Network Posts: 1 Rating: (0)
Suggestion To thank Quote Answer

Share this page:

Share this page on...