APPLICABILITY OF ARTIFICIAL NEURAL NETWORKS TO PREDICT EFFECTIVE THERMAL CONDUCTIVITY OF HIGHLY POROUS METAL FOAMS