Analysis of full-scale membrane filtration processes using econophysics and econometrics

In this paper, econophysics and econometric approaches were employed to investigate statistical and dynamic properties of membrane performance, respectively, using filtration data provided by Woongjin Chemical Co., Ltd., Korea. For the econophysics approach, the filtration index tree was built to comprehensively visualize the closeness between filtration variables, analogous to the hierarchical index tree for the dynamics of stock prices. The logarithmic changes of the filtration variables follow the normal distribution, which implies that the filtration variables themselves are governed by the log-normal distribution due to periodic cleaning and/or seasonal variations. In the econometrics approach, the autoregressive model verified the periodicity of the pressure profile due to cleaning events and natural environmental changes. As developed in this study, the semi-loglinear and autocorrelation models captured pressure growth rates of 1.4% and 0.3 kg/cm2 per week, respectively. Combined and complementary roles of econophysics and econometrics provided new methodological insights to statistically analyze and further forecast membrane filtration performance in terms of the pressure growth rate in the constant flux operation.

Figure optionsDownload high-quality image (20 K)Download as PowerPoint slideResearch highlights▶ The filtration index tree (FIT) comprehensively visualizes the closeness between filtration variables. ▶ Filtration variables (concentration, temperature, flux and pressure) follow the log-normal distribution due to membrane cleaning and seasonal variations. ▶ The autoregressive model captures the periodicity of the pressure profile. ▶ The semi-loglinear and autocorrelation models calculate the pressure growth rates of 1.4% and 0.3 kg/cm2 per week, respectively.