## Fuzzy logic in Scilab (sciFLT). Part 2 — Sugeno

Part 1 (Mamdani) was here.

Let’s solve the same tipping problem as in the previous post but this time using the Sugeno fuzzy inference method.

In the FLS editor, choose

File –>New FLS –> Takagi-Sugeno

Click the ‘Description’ link and fill out the form (click the picture to enlarge).

The input variables (food and service) and the rules are specified in exactly the same way as for the Mamdani system.

See the screenshots of the food and service variables and the rules.

sciFLT can work with either constant or linear Sugeno models.

For a zero-order Sugeno model (constant outputs), the output variable tip can be specified as follows (click to enlarge):

The resulting solution surface is:

Published in: on 27/09/2013 at 11:43  Comments (12)

1. Thank you very much!. I’m having troubles to install the tool box. may you indicate the scilab and sciflt versions you are using. Kind regards from Chile!

• Hi David,

I have Scilab 5.4.1. My sciFLT version is 0.4.4-1

Best wishes!

• It is Running Now!… I think it was something about rights in my computer…

• Great! What kind of task are you solving with fuzzy logic (if it’s not a secret)?

• …we are planning some applications for mining. The idea is to design and do “concept testing” at scilab and then code it at DCS platform… we are thinking in a “trip risk” estimator for a huge conveyor based on motor temp and its ROC. The FIS then may act at the speed high limit, limiting the production to a safe amount… in theory jeje. In another application we are planning to use fuzzy as a kind of suppervisor to control a high rate thickener. The idea is to adjust the discharge rate in terms of torque, level and discharge density (5MFx5MFx5MF=125 rules). A litlle question now: Do you now if there is any restricction for membership functions combination (other than range 0 to 1). Someone told me that they need to sum 1 at any point of the universe of discourse. Is that true?

• They don’t have to sum up to 1 at any point of the universe. They just need to overlap so that there is no point in the universe uncovered by any membership function.

• Date: Fri, 18 Oct 2013 06:10:48 +0000 To: davs001@hotmail.com