Artificial neural networks for the prediction of liquid viscosity, density, heat of vaporization, boiling point and Pitzer's acentric factor Part I. Hydrocarbons

Abstract

A predictive method, based on artificial neural networks (ANN) and equilibrium physical properties, has been developed for the viscosity, density, heat of vaporization, boiling point and Pitzer's acentric factor for pure organic liquid hydrocarbons over a wide range of temperatures (T_reduced≈0.45–0.7). A committee ANN was trained, using ten physicochemical and structural properties combined with absolute temperature as its inputs, to correlate and predict viscosity. A group of 281 compounds, of diverse structure, were arbitrarily ordered into a set of 200 compounds, which were used to train the committee ANN, and a group of 81 compounds, which were used to test the predictive performance of the committee ANN. The viscosity and input data for each individual compound was compiled on average at forty different temperatures, ranging from the melting points to the boiling points for each of the chosen compounds. The mean average absolute deviation in viscosity, predicted by the committee ANN, was ±7.9% which reduces to ±6.5% when the correlated data is also considered. These values are almost a factor of 2 better than other predictive methods and are below the mean average absolute experimental deviation of approximately ±10%, quoted by the DIPPR reference database (AIChE, 1994). In a preliminary study a separate committee ANN was also used to predict the viscosity of the highly polar and hyrdogen bonding compounds, aliphatic acids, alcohols and amines. The predicted mean average absolute deviation for the amines, alcohols and aliphatic acids was ±8.9%. Although this paper deals predominantly with liquid viscosity the same methodology was applied to liquid density, heat of vaporization, boiling point and Pitzer's acentric factor. The predicted mean average absolute deviation for these equilibrium properties was ±0.71%, ±1.04%, ±0.39% and ±5.6% respectively. An attempt has also been made to use the ANN to determine the hierarchical dependencies of viscosity on fundamental molecular and structural parameters.

References

1. Design Institute for Physical Property Data, AIChE, DIPPR, 1994.
2. R. C. Reid, J. M. Prausnitz and B. E. Poling, in The Properties of Gases and Liquids, McGraw-Hill, New York, 4th edn., 1987.
3. J. H. Dymond and M. A. Awan, Int. J. Thermophys., 1989, 10, 941.
4. W. D. Monnery, W. Y. Svrcek and A. K. Mehrotra, Can. J. Chem. Eng., 1995, 73, 3.
5. D. van Velzen, R. L. Cardozo and H. LangenKamp, Ind. Eng. Chem. Fundam., 1972, 11, 20.
6. M. Luckas and K. Lucas, AIChE. J., 1986, 32, 139.
7. P. D. Wasserman, in Advanced Methods in Neural Computing, Van Nostrand Reinhold, New York, 1993.
8. T. J. van der Walt and J. S. J. van Deventer, Chem. Eng. J., 1993, 51, 151.
9. Ming-Jer Lee and Jui-Tang Chen, Ind. Eng. Chem. Res., 1993, 32, 995.
10. Ming-Jer Lee, Shou-Ming Hwang and Jui-Tang Chen, J. Chem. Eng. Jpn., 1994, 27, 749.
11. M. L. Lu, C. McGreavy and E. K. T. Kam, J. Chem. Eng. Jpn., 1997, 30, 412.
12. B. E. Turner, C. L. Costello and P. C. Jurs, J. Chem. Inf. Comput. Sci., 1998, 38, 639.
13. B. E. Mitchell and P. C. Jurs, J. Chem. Inf. Comput. Sci., 1998, 38, 489.
14. B. E. Mitchell and P. C. Jurs, J. Chem. Inf. Comput. Sci., 1997, 37, 538.
15. G. Cybenko, Control, Signals, Sys., 1989, 2, 303.
16. K. Hornik, M. Stinchcombe and H. White, Neural Networks, 1989, 2, 359.
17. S. Haykin, in Neural Networks a Comprehensive Foundation, Macmillan CollegeNew York, 1994, ch. 6.12.
18. R. Chentouf, C. Jutten, M. Maignan and Kanevsky, Nucl. Instrum. Methods Phys. Res. A, 1977, 389, 268.
19. A. Maren, C. Harston and R. Pap, in Handbook of Neural Computing Applications, Academic Press, New York, 1990, p. 238.
20. C. M. Bishop, in Neural Networks for Pattern Recognition, Clarendon Press, Oxford, 1995, 364.
21. O. Ivanciuc, J. P. Rabine, D. CabrolBass, A. Panaye and J. P. Doucet, J. Chem. Inf. Comput. Sci., 1996, 36, 644.
22. D. Cherqaoui and D. Villemin, J. Chem. Soc., Faraday Trans., 1994, 90, 97.
23. M. T. Hagan and M. Menhaj, IEEE Trans. Neural Networks, 1994, 5, 989.
24. C. M. Bishop, in Neural Networks for Pattern Recognition, Clarendon Press, Oxford, 1995, p. 372.
25. D. J. C. MacKay, Neural Comput., 1992, 4, 448.
26. D. J. C. MacKay, Neural Comput., 1992, 4, 415.
27. H. G. Yucel and S. Ozdogan, Can. J. Chem. Eng., 1988, 76, 148.
28. D. van Velzen, R. L. Cardozo and H. LangenKamp, Euratom, 4735e, Joint Nuclear Research Centre, Ispra Establishment, Italy, 1972.
29. C. R. Duhne, Chem. Eng., 1979, 86(15, Part 1), 83.
30. C. L. Yaws, X. Lin and L. Bu, Chem. Eng., 1994, 101, 119.