Intelligent approaches for linear controllers tuning with application to temperature control

Abstract

Despite the new sophisticated controllers the linear PI/PID controllers keep the leading position in industrial implementations due to their simplicity, easy and well developed design and tuning, good system performance and robustness and universal application area. Various enhancements have been suggested to enlarge their range of operation with good performance. Fuzzy logic (FL) and genetic algorithms (GAs) offer proper intelligent solutions for improving the PI/PID controllers auto-tuning and adaptation to successfully deal with plant nonlinearity, inertia and changing parameters without plant model. The aim of this research is to develop a simple for engineering use method for design of a fuzzy two-level controller (FTLC) of a linear PI/PID controller and a FL supervisor. The FL supervisor tunes on-line the scaling of the basic PI/PID controller's gains depending on performance indicators. The basic FTLC parameters are off-line optimized using GAs and a proposed fitness function of system performance and energy efficiency which is estimated in system simulation with a GA developed T-S plant model. The method is tested in real time control of the temperature in a laboratory -scale dryer. The result is decreased settling time, overshoot and energy consumption in the whole range of operation of the nonlinear plant.

Disturbance Compensation in Fuzzy Logic Control of Level in Carbonisation Column for Soda Production

“…It is preferred because it is precise (free of industrial noise and disturbances), fast and saf e for the industrial plant. GA optinisation is applied basically for TSK plant modelling and FLC parameters tunin g [6,7]. In order to ensu re mapping of the system nonlinearity and adaptive abilities it is important a correct fitness function to be defined and the data used to be rich in frequencies and magnitudes, i.e.…”

“…The model-free fuzzy logic controllers (FLC) [1,2] and the based on Takagi-Sugeno-Kang (TSK) plant model fuzzy parallel distributed compensation (PDC) [3,4 ] enable an unified design of various nonlinear controllers that ensure s table, high performance, energy-efficient and robust plant control. The TSK plant model, the FLC and the PDC parameters are tuned b y the hel p of off-line optimisation based mainly on genetic algorithm s (GA) and experimental or simulation data [5,6,7].…”

“…FLC and fuzzy gain schedulers are devel oped for level control in boilers, nuclear generators, tanks, etc. [6,10,11 ]. They are tested via simulations [12] or in real time control [ 13,14,15].…”

Intelligent Approaches to Real Time Level Control

“…The present work is based on the developed design procedures for off-line GA tuning of linear controllers [21] and local linear controllers in a PDC used with a FL supervisor (FLS) [22] applied for temperature real time control. There it is established in real time control that the effect of GA optimization of the main linear controller is equivalent to using of an on-line FL supervisor.…”

“…3 and is based on the structure, suggested in [21]. It assumes that the plant in the closed loop system operates in three overlapping linear domains for all possible reference changes and disturbances and the dynamics of the local linear plants can be represented by a series connection of two time-lags, the second is common for the three parallel channels, unlike the dynamic models in [21]. The TSK plant model consists of a zero-order Sugeno Model with fuzzy rules:…”