| RESEARCH | RESEARCH PROJECTS | Soft computing methods and techniques Artificial neural networks: structure optimization of multi-layer feedforward networks; GMDH (Group Method and Data Hadling) networks and their extentions; networks with the dynamic model of neurons; neural network ensembles; training algorithms: gradient and genetic fuzzy and neuro-fuzzy systems: structure and parameter optimization via evolutionary and gradient algorithms; bounded-error analysis expert systems: integrated knowledge bases, knowledge representation (rules, analytical, neural), knowledge aquisition, fuzzy logic; applications to air pollution processes and power plants diagnosis systems Fault detection and isolation (FDI) Analytical methods: model-based approaches, robust observers (Kalman and Luenberger filters), unknown input observers (bounded-error technique, mathematical model design via genetic programming) Soft computing methods: neural and neuro-fuzzy models and classifiers, optimization of diagnostic systems via evolutionary algorithms, expert systems, integration of qualitative and quantitative knowledge Modelling and simulation Distributed parameter systems: modelling, state and parameter estimation, Kalman filters, sensors and actuators location (stationary and moving) Applications: air pollution processes on the city scale; nuclear power reactors, industrial processes -- sugar and power plants