Search / Korean Journal of Chemical Engineering
HWAHAK KONGHAK,
Vol.31, No.6, 744-752, 1993
신경회로망을 이용한 Joback Group Contribution Method의 개선
Modification of the Joback Group Contribution Method Using Artificial Neural Networks
인공지능의 기법을 활용하여 Group Contribution Method의 정확도를 향상시키는 일반적인 방법을 제안하였다. Joback GCM을 선택하고, 순수한 물질 400여개의 데이터를 사용하여 역전파 신경회로망을 30000-40000회 학습시켰다. 학습된 가중치 매트릭스를 사용하여 임계성질, 끊는 점, 어는 점, 생성엔탈피, 생성 Gibbs 자유에너지 및 이상기체의 비열을 예측해 본 결과, 기존의 수학적 모델을 이용하여 개발된 GCM보다 정확한 예측을 보임을 확인하였다.
Using the technique of artificial intelligence, the general method to improve the accuracy of Group Contribution Methods was proposed. The Joback GCM was chosen and the back propagation neural networks were trained 30000 to 40000 times with the experimental data of more than 400pure substances. The estimations of the critical properties, the normal boiling and the freezing temperatures, heats of formation, Gibbs free energies of formation, and ideal gas state heat capacities using trained weight matrices were much better than those from the conventional GCM developed on the basis of an assumed mathematical model.
[References]
  1. Sherwood TK, Pigford RL, Wilke CR, "Mass Transfer," McGraw-Hill, New York, NY, 1975
  2. Gani R, Brignole EA, Fluid Phase Equilib., 13, 331, 1983
  3. Joback KG, Stepanopoulos G, Siirola JJ, Grossmann IE, Stephanopoulos G, "Foundations of Computer-Aided Process Design," Elsevier, Amsterdam, 1990
  4. Nielsen B, Gani R, Fredenslund A, Bussemaker H, Ledema PD, "Computer Applications in Chemical Engineering," Elsevier, Amsterdam, 1990
  5. Gani R, Nielsen B, Fredenslund A, AIChE J., 37(9), 1318, 1991
  6. Jung JY, M.S. Thesis, Postech, Pohang, Korea, 1993
  7. Klincewicz KM, Reid RC, AIChE J., 30(1), 137, 1984
  8. Joback KG, Reid RC, Chem. Eng. Commun., 57, 233, 1987
  9. Lapedes A, Farber R, "How Neural Nets Works," Neural Information Processing Systems, Denver, CO, 1987, American Institute of PHysics, 1988
  10. Bhat N, McAvoy TJ, Comput. Chem. Eng., 14(4-5), 573, 1990
  11. Reid RC, Prausnitz JM, Poling BE, "The Properties of Gases and Liquids," 4th ed., McGraw-Hill, New York, NY, 1987
  12. Freeman JA, Skapura DM, "Neural Network Algorithms, Applications and Programming Techniques," Addison-Wesley, Reading, MA, 1991