The dissolved oxygen ramp is immoral

Abstract

On-site recovery and reuse allow saving dwindling water resources. However, for safe water reuse, it is essential to be able to guarantee a given water quality at all times. This is especially challenging for on-site water resource recovery facilities because sensors are unmaintained. Therefore, robust features for monitoring the treatment process are required. Here we present a method, i.e. Dynamical Systems Analysis (DSA), to explore models for informative features and systematically characterise these features. We exemplarily demonstrate the application of DSA to the Activated Sludge Model 1, and explore a ramp feature in the dissolved oxygen signal used to detect full ammonium oxidation. Our findings show that the ramp feature is immoral, which means multiple state variables cause the observation of this feature, potentially leading to false positive detections. However, through DSA, we show that the prediction accuracy is increased when the feature is selected based on a minimal value of the slope of the dissolved oxygen signal. The current analyses also provided further features that could increase the accuracy of the detection, in particular the concentration of autotrophic bacteria. We further highlight that the potential of DSA goes far beyond our results, including feature identification, calibration procedure and experimental design. We additionally provide all code open access and with examples so that readers can conduct their own analysis.