Abstract

In industrial rice processing plants, Boilers plays a crucial role in generating steam for parboiling of rice paddy, hence it requires a precise regulation. Conventional controllers are commonly used but struggle with stability during pressure changes, load variations, and emergencies due to system nonlinearity. Traditional control methods are often ineffective in such a complex system. This paper develops a model of a husk ¬fired boiler system used in rice processing plant with two inputs (husk fired mass flow rates 1 and 2) and two outputs (steam pressure and water level) using Artificial Neural Network (ANN). The output parameters of the plant are predicted using direct data collected from the actual boiler plant. A Nonlinear Autoregressive with External (Exogenous) input (NARX) neural network which has linear output neuron (purelin) and Hyperbolic tangent sigmoid transfer function (Tansigmoid) was used at the output and hidden layer respectively. Levenberg-Marquardt back propagation Algorithm (trainlm) is then used to train the NARX neural network. With mean squared error (MSE) of 0.0077 and regression value (R) of 0.99, the developed neural network model showed that the simulated model outputs closely match the actual boiler plant outputs when tested using the validation input data. Moreover, fuzzy logic controller (FLC) was implemented to control the identified boiler model. Simulation results demonstrated that the water level control with fuzzy controller achieved better response with settling time of 1.32 sec and steady state error of 5.41. This is against the PI controlled model where the settling time of 1.73 sec and steady state error of 8.11 are recorded. The large steady state error observed for the water level control might result from the limitation in the collected data. During model identification the data for water feed rate and steam flow which has direct influence on water level is not available, hence the water level is controlled indirectly through husk feed rate which influences both steam pressure and water level. On comparison with a PI controlled boiler system, the steam pressure control using fuzzy controller achieved good response by eliminating overshoot and settling at 4.53 sec, which is better than the PI controlled case having an overshoot of 2.35% and settling time of 5.12 sec. These performance improvements highlighted the technical significance of the FLC in handling nonlinear and disturbance- prone boiler dynamics, ensuring more stable regulation, enhanced robustness, improved operational safety, and higher thermal efficiency compared to conventional PI control in rice processing applications.

Key words: Model Identification; Artificial neural network (ANN); Husk fired Boiler; PI Controller; Fuzzy Logic Controller;