10.6 Review/Preview

Empirical computer models play a key role in a variety of industries for tackling both design and control problems. Generally, the calibration or tuning of these models, the selection of values for the empirical constants, is a time-consuming task. Fortunately, the search capabilities of genetic algorithms can be used to help solve the problem of selecting these constants.

There are numerous criteria for judging accuracy of empirical models. The traditional least squares criteria is in widespread use, but it does not perform well in the presence of outliers. Thus, a more robust technique, least median squares, has been presented. The LMS method works well when used in conjunction with genetic algorithms to produce regression-based computer models based on data obtained from physical systems. The effectiveness of this approach to LMS computer modeling has been demonstrated on several systems including a hydrocyclone, a grinding circuit, and a column flotation circuit.

One of the major drawbacks of empirical models is that the form of a modeling equation must be known in advance. Such an equation is not always available, and thus alternative techniques must be considered. Frequently, data collected from a plant environment is the only information available to the modeler. In such a situation, regression techniques are extremely difficult to develop, as an effective equation form must be discovered. Alternatively, the modeler can employ techniques that are capable of modeling problems using only the available data. Such a technique is presented in the next chapter. Neural networks are computational paradigms of the human brain and their associated theory has been developed by researchers from the field of artificial intelligence. As will be seen in the following chapter, neural networks can be used to develop accurate computer models.

References

Agar, G. E., Huls, B. J., & Hyma, D. B. (Eds.) (1991). Column ’91. The Canadian Institute of Mining and Metallurgy and Petroleum.

Alford, R. A. Modelling and design of flotation column circuits. Doctoral dissertation, Queensland, Australia: The University of Queensland, 1991.

Finch, J. A., and Doby, G. S. Column flotation. Toronto: Pergamon Press, 1990.

Herbst, J. A., and Rajamani, K. Models for the dynamic optimization of mineral processing plant performance. In K. V. S. Sastry and M. C. Fuerstenau (Eds.) Challenges in mineral processing. SME, 709-739, 1989.

Huber, P. J. (1981). Robust statistics. New York, NY: John Wiley & Sons.

Karr, C. L, Weck, B., Massart, D. L., & Vankeerberghen, P. (1995). Least median squares curve fitting using a genetic algorithm. Engineering Applications of Artificial Intelligence, 8(2), 177–189.

Karr, C. L. Optimization of a computer model of a grinding process using genetic algorithms. In H. El-Shall, B. Moudgil, and R. Weigel (Eds.), Beneficiation of Phosphate: Theory and Practice (pp. 339–345). Littleton, CO: Society for Mining, Metallurgy, and Exploration, Inc., 1993.

Karr, C. L., Stanley, D. A., & Scheiner, B. J. (1991). A genetic algorithm applied to least squares curve fitting (Report of Investigations number 9339). Washington, DC: U.S. Department of the Interior, Bureau of Mines.

Luttrell, G. H., Adel, G. T., and Yoon, R. H. Modeling of column flotation. SME Annual Meeting, Preprint number 87-130, 1987.

Mathews, J. H. (1992). Numerical methods for mathematics, science, and engineering. Englewood Cliffs, NJ: Prentice Hall.

Mehta, R. K., Kumar, K. K., and Schultz, C. W. Multiple objective optimization of a coal grinding process via simple genetic algorithm, Society for Mining, Metallurgy, and Exploration, Inc., Littleton, CO, preprint number 92-108, 1982.

Plackett, R. L. (1972). Studies in the history of probability and statistics XXIX: The discovery of the method of least squares, Biometrika, 59, 239–251.

Plitt, L. R. A mathematical model of the hydrocyclone classifier. CIM Bulletin, 69: 114–123, 1976.

Rousseeuw, P. J. (1984). Least median of squares regression, Journal of the American Statistical Association, 79, 871–880.

Rousseeuw, P. J., and Leroy, A. M. (1987). Robust regression and outlier detection. New York: John Wiley & Sons.

Sastry, K. V. S. (1978). Beneficiation of mineral fines, problems and research needs. Report of a workshop held at Sterling Forest, New York.

Ynchausti, R. A., Herbst, J. A., and Hales, L. B. Unique problems and opportunities associated with automation of column flotation cells. Column Flotation ’88, SME, 27–33, 1988.