Abstract
Although follow-fiber membranes have been used in many membrane water treatment plants, they are subject to membrane failure during the course of practical usage. Thus, it is important to assess the ageing of follow-fiber membranes for replacement of membrane modules. Therefore, this study aimed to predict the numbers of follow-fiber membrane failure and to estimate appropriate exchange periods of membrane modules. Membrane failure data were obtained from a full-scale plant for the analysis by various statistical models. The acceptable level of log reduction value (LRV) was set at 2.26 based on the raw water quality data of the membrane water treatment plants in Japan and the desirable water quality after membrane filtration. The permissible membrane failure rate was then estimated to be 0.028% to maintain a higher LRV than 2.26. However, based on the water flow model simulation within a membrane module with failed hollow-fiber membrane, it was also found that the LRV of the membrane module is highly dependent on the location of membrane fiber failure. It was demonstrated that both linear model and generalized linear model using logistic regression fit well with the the membrane failure data of a full-scale plant, and can be used to estimate the periods for membrane module replacement. However, binomial distribution model was found to be not suitable for analysis of membrane failure data.
