Model selection for prediction and interpretation Biosensor development and signal processing for real-time water quality monitoring PhD student: Dolores Gonzalez. Supervisors: Mirella Di Lorenzo, Petra Cameron. Dept. of Chemical Engineering, University of Bath, BA2 7AY, UK. E-mail: dg568@bath.ac.uk Problem statement Principles of MFC Biosensors The quality of the water bodies is decreasing due to agricultural runoff of pesticides and herbicides. Currently, bioassays are used to evaluate the effect of these compounds on biodiversity. These methods are insufficient to provide a continuous, real-time measurement of the water quality. There is a need to develop an online, continuous and real time sensors for water quality monitoring. A microbial fuel cell (MFC) is an electrochemical device that converts chemical energy into electricity by using the metabolic activity of micro-organisms, Fig.1. As electricity generation depends on the bacterial metabolism, which is affected by any disturbances (i.e. the present of a pollutant), a correlation between the output signal and the target could be established. MFC biosensors could provide a cost-effective solution to these problems, however, a research gap in signal processing has to be addressed before the technology can be commercialised. Runoff of pollutants Biotoxicity Project aim The aim of this project is to improve the knowledge in signal processing of MFC biosensors to promote practical applications. [1] Objectives To design a continuous and real-time MFC biosensor. To prove the ability of the sensor to detect components of interest. To develop a model to interpret and predict the sensor’s output. To validate the sensor in the field. [2] Research methods Fig. 1 Basic principle of a MFC biosensor. Data Objective Model approach Procedure Previous data Literature review MFC sensor Data generation Detection of pollutants Regression Suitable for one factor at the time, in controlled conditions. Different toxicant concentrations Model calibration and validation [5] Model screening [5] Optimum operational conditions ANOVA/ RSM Suitable to assess interactions and visualise optimum operational conditions. Model selection for prediction and interpretation [4] [5] Machine learning Prediction of complex patters. The functions have no physical meaning. Interpretation of complex patterns References [1] Sediment runoff from a farm in Iowa (2011). By Lynn Betts / Photo courtesy of USDA Nat. Res. Cons. Service. [2] Chouler, J. & Di Lorenzo, M. Biosensors 5, 450–470 (2015). DOI:10.3390/bios5030450 [3] Feng, Y. & Harper, W. F.. J. Environ. Manage. 130, 369–374 (2013). DOI: 10.1016/j.jenvman.2013.09.011. [4] Hosseinpour et al. Journal of Environmental Health Science & Engineering 2014 12:33. DOI:10.1186/2052-336X-12-33 [5] Chouler J. Development of a self-sustainable and cost-effective tool for water quality monitoring in developing countries. PhD Confirmation report. CSCT University of Bath 2015. Optimisation algorithms Improve the efficiency in determining parameters when the system is undetermined. [3]