Suboptimal Recursive Methodology for Takagi-Sugeno Fuzzy Models Identification - Fuzzy Modeling and Control: Theory and Applications - page 36

25

20

15

10

5

0

0 50 100 150 200 250 300 350

Iteration

30

25

20

15

10

5

0

0 50 100 150 200 250 300 350

Iteration

Fig. 2.4 Final models errors from training data

1 1 1 16

0 50 100 150 200

Iteration

1 1 1 16

0 50 100 150 200

Iteration

Fig. 2.5 Final models errors from validation data

Algorithm EKF perform better because it can adjust the antecedents and this

gives it more flexibility. Analyzing the standard deviations in Table 2.1 , is possible
to see that the algorithms are fairly consistent, since they get very similar results to
changes in input data and noise that affects the equation.

( c + a )

2.4.1.2 Case II: Membership Functions of Different Type
and Antecedents Restriction

In this case, the antecedents of the initial TS fuzzy model are defined by a S member-
ship function, a Z , two trapezoidal and two triangular. These antecedents have been