A Comparative Study for Identification of Distributed Parameter Systems using Sensor Networks and Adaptive-Network-Based Fuzzy Inference

The paper presents a study upon the possibility to use adaptive-network-based fuzzy inference method (ANFIS) in the identification of distributed parameter systems, implementing a distributed sensor network in the system. Some main properties of different identification methods are presented with possible application. The fuzzy systems, implemented using rule bases, fuzzy values, membership functions, fuzzyfication and defuzzification methods, may be used to describe distributed parameters systems. Also the feedforward neural network is a good choice. A combined method is the adaptive-network-based fuzzy inference, which implement the fuzzy system as a near network trained to learn the model of distributed parameter system. A study case of a heat transfer system is considered. Models as differential equations and approximation of these equations are considered. Meshes, isotherms and temperature estimate values are presented for different numbers of sensor nodes placed in heat transfer space. The attentions is focused on the number and positions of the sensor nodes in the distributed parameter system to assure good accuracy of the estimates. Key-Words: Distributed parameter systems, system identification, modeling of dynamic systems, fuzzy logic, neural networks, sensor networks, artificial intelligence, neuro-fuzzy modeling, adaptive network based fuzzy inference, heat transfer equation, partial differential equations, mesh generation, finite element methods, modeling in science and engineering.