Applying reliability analysis to evaluate water treatment plants

The aim of this study was to demonstrate the applicability of reliability analysis to a water treatment plant (WTP). Data from daily monitoring of raw and finished water quality from a direct filtration WTP covered the parameters of turbidity, apparent colour, pH, aluminum, chlorine and fluoride, from December 2007 to August 2011. Data analysis included descriptive statistics and adherence tests to normal, log-normal, rectangular, exponential and gamma distributions. A reliability analysis was conducted in three steps: (i) estimation of mean values for monitored parameters based on the established percentage of compliance and comparison with observed mean; (ii) estimation of expected percentage of compliance based on observed mean values and comparison with the observed percentages; and (iii) evaluation of the applicability of reliability analysis to parameters with upper and lower limits. When the comparison between estimated and observed percentages of compliance was made, reliability analysis led to inaccurate results for parameters whose observed percentage of compliance remained below 90%. For parameters whose observed percentage of compliance tended to fullness, deviations were not observed. It is believed that such a conclusion can be extrapolated to estimated mean values: most accurate results are obtained for parameters that have percentages of compliance near 100%.