Yuliati, Hadi Santosa
The Waste Heat Boiler (WHB) System is a steam-generating heat system that harnesses the energy transfer from combustion to transform water into steam, serving as the primary output. Identifying the plant model of the WHB unit poses challenges due to its dynamic characteristics and nonlinear behaviour, which are susceptible to the influence of numerous factors. This study involves the implementation of experiments aimed at identifying and collecting input data are boiler feed water flow and steam product flow, and its output is the real-time water level in high pressure drum from the operational activities of a fertilizer factory located in Gresik, East Java. The application of the Radial Basis Function-Neural Network (RBFNN) identifier is justified in identifying the dynamics of a WHB system due to its favourable approximation properties and straightforward topological structure. The optimal topology for the single hidden layer structure RBFNN model with learning rate 0.5 involves using five identifier input vectors and nine RBF nodes in the hidden layer, resulting Normalized Square Root Mean Square Error (NSRMSE) values of 0.2299. The minimal final prediction error also indicates indicate that the output of the RBFNN model is capable of accurately estimating the dynamic characteristics of actual measurements. © School of Engineering, Taylor’s University.
Electrical Engineering Study Program, Widya Mandala Surabaya Catholic University, Jalan Kalijudan 37, Surabaya, East Java, Indonesia; Engineer Professional Study Program, Widya Mandala Surabaya Catholic University, Jalan Kalijudan 37, Surabaya, East Java, Indonesia