Stochastic Experimentation Principles and Practice

• Published in: Journal of the Royal Statistical Society. Series D (The Statistician) 1967-01, Vol.17 (2), p.157-181
• Main Author: Shelton, R. J.
• Format: Article
• Language: English
• Subjects: Autocorrelation; Binary codes; Flow velocity; Fourier transformations; Frequency ranges; Nyquist frequencies; Quick response codes; Signal noise; Spectral energy distribution; Spectroscopy; Statistics at the BP Research Centre
• ISSN: 0039-0526; 1467-9884
• DOI: 10.2307/2986897

Considerable experience has been gained in the past in building steady‐state models of refinery processes. Recently interest has been aroused in employing these models for process control using computers. In such applications plant dynamics become important. This paper outlines the theoretical basis for the application of spectral analysis techniques to the dynamic characterization of an operating process. Spectral analysis is particularly powerful when used in conjunction with random perturbing signals. The properties and methods of generation of two such signals, the random telegraph signal and the pseudo‐random binary chain code, are discussed. Simple examples of the pseudo‐random binary chain code are given and a practical method of generation is included. Previous workers have used pseudo‐random binary chain codes for time domain characterization but the present paper shows that it is especially valuable in the frequency domain. The techniques outlined have only been applied by other workers in computer simulations of processes or pilot plant studies. The present report illustrates the description of the methods with an account of recent practical work on a distillation column at BP Refinery, Llandarcy.