Department Process Engineering

Project: SoDAN

Soft-sensing, Diagnosis and Automation for Nutrient Recovery


Centralised sewer-based wastewater management is too costly for most cities in the world. An attractive alternative consists of a network of reliable small-scale reactors which treat the different wastewater streams according to their properties. Urine is the wastewater stream with the highest nutrient content and several processes such as ammonia stripping, struvite precipitation, and nitrification/distillation have been studied and tested in pilot-scale. However, all of these processes have been operated only under close scientific supervision; Long-term operation under realistic conditions remains a challenge. For instance, the biological nitrification process used to stabilise urine and avoid losses of ammonia through volatilisation is sensitive to rather common deviations from nominal load conditions (flow rate and concentration changes) as well as equipment failures (e.g., air/oxygen supply). Since this can lead to a persistent failure of the nitrification process, it is desired to develop automated methods for operation of such plants so that high process efficiencies can be reached under normal conditions while also offering sufficient flexibility to handle anomalous events (load changes, equipment failures) in a timely fashion. This project aims at developing an automated control system for a urine nitrifying plant. The control system will consist of a top-level supervisory controller which is based on the use of data-driven models (i.e. models built entirely on previously measured data) for process diagnosis and a bottom-level control strategy based on a first principles model (i.e., built from fundamental, mechanistic process knowledge). To this end, the project will investigate and compare (1) methods for data dimensionality reduction, (2) data-driven process diagnosis methods, and (3) model based observers and (4) different model based control formulations. Each element of the control system will be developed and tested first in a simulated environment, using a detailed model of the plant as obtained through previous studies. In a second step, these elements will be tested on a full-scale experimental setup, which is already installed at Eawag. Testing on a full-scale reactor will permit to improve and fine-tune the developed methods. An automation system which combines first principles mechanistic models with data-driven models will allow stable process operation not only under normal conditions, but also in anomalous situations. We are convinced that this automation system will facilitate reliable operation of complex biological processes, including small decentralised reactors. Such reactors can be used to establish a sanitation system that adapts more easily to demographic and environmental changes than large sewer-based wastewater management systems.


Funding:Swiss National Science Foundation (SNF)
Spike staff:Christian Thürlimann