On-line estimation of the aerobic phase length for partial nitrification processes in SBR based on features extraction and SVM classification

•We propose a strategy for on-line estimation of the nitrification phase duration.•Estimation of nitrification phase duration depends on real-time measurements of both pH and dissolved oxygen.•We use feature extraction and support vector machines to solve the on-line estimation problem.•A database of 533 SBR cycles was used to train and test the proposed strategy.•We achieve 100% correct classification for both training and testing databases.

We present a strategy for the on-line estimation of the aerobic reaction phase length for a partial nitrification process with pH and dissolved oxygen closed-loop control. To overcome existing drawbacks associated to partial nitrification (e.g., non-linearities and time-variant behaviors), our strategy is based on feature extraction over manipulated variables to identify interesting patterns associated to the end-point of nitrification. We use a support vector machine (SVM) classifier as a decision tool to determine the end-point of the aerobic phase. A database of lab-scale sequencing batch reactor (SBR) cycles selected from ten months of operation was used to train and test the proposed decision-making strategy. Results for all 533 SBR cycles showed 100% correct classifications. Most aerobic phase lengths in the analyzed database had a reduction time around 20 min, although time reductions greater than 60 min were also achieved.