30
Prediction of Mechanical Property of Compounds by Using a Genetic Algorithm and Artificial Neural Network
TECK CHENG, SENG*
Engineering, Design & Prototype Centre, Centre of Excellence,
Malaysian Rubber Board, RRIM Research Station,
47000 Sg. Buloh, Selangor D.E., Malaysia
The ability of a Genetic Algorithm (GA) and artificial neural network to evaluate the dynamic shear stiffness of rubber compounds from their formulation is presented. An orthogonal experimental design (OED) was used to explain the effect of formulation, processing and testing conditions on dynamic shear stiffness of rubber compounds. The OED method was adopted because of its benefit that it requires minimum number of experiments to represent a large number of experimental data. In this work a three-layer network with GA and back-propagation (BP) learning algorithm was developed and trained. The number of hidden nodes was determined to be 16. Results from the GA-BP network were compared with experimental data. The GA-BP network was found to generalize well at shear strains higher than 10%.