Jurnal Online - http://www.jurnal.lipi.go.id


Noor Fachrizal

UPT-LSDE BPPT, Puspiptek Serpong, Tangerang

Mathematical model of drying process is non-linear, by linearization the model become time-varying. In order to apply a control system, the non-stationary model of the process must be estimated during drying process. A recursive extended least square estimation algorithm with improved adaptation gain using variable forgetting factor is proposed to identify model of the system. Validation of the model is based on statistic testing criteria : FPE, Whiteness test and Chi Squared test. Estimation algorithm is tested off-line by using each minute sampled data of Meranti's drying process of Solar Wood Drying System's running. Control signals of the model are optimized using linear quadratic criteria function, with Liapunov approach. By trial and error method for finding weighting function, it results optimal control laws for enforcing moisture content of wood following its set-point. Dynamic model of the process considered both collector and boiler supply as a unity because of unpredictable solar energy intensity. In fact, both these actuators are separated physically. Therefore, it needs an algorithm to maximize solar energy utilisation. This regulation depends on temperature difference of collector and drying chamber, and control signal from controller. This regulation rule is non-linear, it is difficult to determine model parameter explicitly. Artificial Neural Network (ANN) model gives possibility to formulate the model implicitly.

Publication : Physics Journal IPS A5 (2003) 0218
Full paper : format PDF (139.447 byte)
Contact : author(s)
Additional info : 4 pages, language English