A detailed survey of the field of intelligent control is presented. Current practices are reviewed and the need for a unifying framework to identify and strengthen the underlying core principles is postulated. Intelligent control is redefined to make explicit use of human systems in control as a reference model. Psychological theories of intelligent behaviour reveal certain basic attributes. From these a set of necessary and sufficient conditions for intelligent control are derived. Learning ability is identified as a crucial element. Necessary attributes for learning are prediction capabilities, internal world model, estimation of the model parameters, and active probing to reduce uncertainties. This framewoik is used to define a Learning Based Predictive Control (LBPC) strategy. LBPC is derived from Predictive Functional Control techniques with an adaptive layer implemented by recursive least squares. Improved performance above conventional adaptive control is demonstrated. Distributed parameter systems are identified as a suitable application area requiring an intelligent control approach. Such systems are invariably complex, ill-defined, and nonlinear. Plasticating extrusion processes are considered in particular. LBPC is applied to control of the primary loop to regulate melt temperature and pressure at the die. A novel control technique is proposed for dynamic profile control of extruder barrel wall temperature. This is a two-level hierarchical scheme combining the benefits of LBPC control blocks at the lowest level with decision logic operating at the higher level as a supervisor. This Logic Based Strategy allows multivariable control of non-square systems with more outputs than inputs. The application of LBS to an extruder is demonstrated.