Applicability of statistical analysis for performance and reliability evaluation of large-scale water treatment plants with direct filtration systems.

Water treatment plants (WTPs) are extremely important in basic sanitation services because of their association with human health by producing safe drinking water. Thus, their proper operation is of utmost relevance and has led to the development of distinct performance evaluation methodologies. Direct filtration is a leading technology applied in WTPs. However, although it costs less than conventional treatment, it might also be less flexible and robust. To evaluate performance with data from real-scale WTPs with direct filtration systems, the use of a statistical analysis methodology for turbidity, apparent color, and pH data from raw and effluent water is proposed. Reliability analysis, a probabilistic-based methodology, was applied for turbidity alongside evaluating the compliance of the treated water with different potability standards. The parameters pH (between 6.0 and 8.0) and apparent color (< 15 HU) showed almost complete compliance (> 99%). Reliability analysis could not be applied for apparent color and pH, while apparent color did not adhere to the lognormal distribution frequency. The turbidity results show high variability in the coefficients of variation and reliability among various plants and a general difficulty in complying with stricter standards, such as the 0.1 and 0.3 NTU defined by the United States Environmental Protection Agency. All WTP studies show a 95% compliance with the World Health Organization standard of 5.0 NTU. A higher value for the coefficient of reliability and a lower value for the variation coefficient could indicate a more stable process, regardless of the treated water quality, highlighting the relevance of applying combined methods for performance evaluation, such as compliance with established standards.