【正文】 ngs together small degradation of dynamical behavior (Fig. 7). This degradation can be almost eliminated by fine tuning of the parameters. As a result of these simulations we can state that the assertions in paper [6] were not proved. 3. FUZZY PD+PI CONTROLLER DESIGN 9 Let us create fuzzy controller as a parallel bination of fuzzy PI and PD controllers. Simulation results obtained using classical IPD controller are shown in Fig. 9. This controller was adjusted using ZieglerNichols method (K = , TI = s, TD = s). The initial adjustment of fuzzy PIPD controller is taken up from reference [2] (KI = KD = 4, TI = , TD = 2, MI = MD = 10, T = s, MinMax inference method, COG method for defuzzification, 7 membership functions). Membership function layout and rule base for both PI and PD parts are shown in Fig. 1 and Fig. 2 respectively. The only exception is in increment of mand which is realized as shown in Fig. 11. Simulation results can be seen in Fig. 10. Fig. 12 shows the time responses with three different settings. Solid line uses the same adjustment as previously described simulation but inference method is ProdMax. Dashdot dot line represents the simulation with 3 membership functions, inference MinMax and defuzzification using triangular membership functions with COG method. Dashanddot line shows simulation results with 3 membership functions, inference MinMax and defuzzification using singletons. When the singletons are used they are placed into the vertex of original fuzzy membership functions. The influence of disturbance with small amplitude of which acts at the input of the system is shown in Fig. 13. 10 Fuzzy controller is generally inclinable to oscillation with relatively small amplitude. The origin of this oscillation is not only incorrect tuning of the parameters but also the inference method MinMax. The oscillations are considerably eliminated when the ProdMax inference method is employed and singletons for defuzzification are used. Another potential source of oscillations is wrong implementation of inference engine. Shift of the vertex point of middle membership function just for on the normalized universe (illustrated in Fig. 15) causes the limit cycles to appear (see results in Fig. 14) without any other external intervention. The initial deviation is caused by PD controller and the oscillations by PI controller. When the singleton membership functions are used instead of triangular ones the described phenomena of oscillations disappear. Following the results in Fig. 14 it can be seen that the singletons as an output membership functions give at the beginning higher oscillation but after the time they disappear (dashed line). Triangular membership functions with COG defuzzification give after short transient response steady limit cycle. Note the nonzero steady state error in both cases. As a consequence we can say that there is a difference between defuzzification using singletons and tria