Statistical Monitoring of Water Systems

In this work a new statistical control methodology that works in the original variable space is presented. The study is focused on the performance of the detection stage, which applies the 2-order Mahalanobis distance as statistical test. Its critical-value is estimated using a kernel density estimation technique. The identification stage uses the concept of the nearest in control neighbor to calculate the variable contributions to the inflated statistic and employs the same distance measure. Application results are provided for a waste water treatment plant.