15 th International Conference on Environmental Science and Technology Rhodes, Greece, 31 August to 2 September 2017 CEST2017_00700 Electronic nose performance optimization for continuous odour monitoring in ambient air Zarra T. 1 , Cımatorıbus C. 2* , Naddeo V. 1 , Reıser M. 3 , Belgıorno V. 1 And Kranert M. 3 1 Sanitary Environmental Engineering Division (SEED), Department of Civil Engineering, Università degli Studi di Salerno, via Giovanni Paolo II, 132 - 84084 Fisciano (SA), IT 2 Faculty of Building Services Energy Environment, University of Applied Sciences – Hochschule Esslingen, Kanalstraße, 33 - 73728 Esslingen, D 3 ISWA Institut, Department of Civil Engineering, Stuttgart University, Bandtäle 1,2, 70569 Stuttgart, D *corresponding author: e-mail: Carla.Cimatoribus@hs-esslingen.de Abstract Industrial plants with odour emissions affect the quality of air and are often cause of public complaints by the people living surrounding the plant. For this reason, the control of odour represent a key issue. The starting point for an effective odour control it‟s their objective quantification. The electronic nose represent the odour measurement technique with probably the greatest potential, but currently there is not a universally recognized procedure of their application for the continuous monitoring of environmental odours.The aim of this paper is to present and describe a novel procedure to training electronic noses in order to maximize their capability of operating a qualitative classification and estimating the odour concentration of ambient air. This novel approach will reduce the uncertainty and increase the reliability of the continuous odour measures. The research is carried out through a real case study application in a big liquid waste treatment plant (LWTP). The seedOA system, patented by the SEED group of the University of Salerno, was used as e.nose device. The characterization of the odour concentrations from the different treatment units and the identification of the principal odour sources is discussed. Keywords: air quality, dynamic olfactometry, liquid waste treatment plant, multisensory array system, public complaints. 1. Introduction Offensive odour emitted in ambient air from different types of industrial plants are among the main causes of conflict by the community living surrounding the plants and of complaints at the local authorities (Zarra et al., 2008; Belgiorno et al., 2012). A prolonged exposure to odours causes a variety of undesiderable reactions in people, including unease, headaches, respiratory problems, nausea or vomiting (Zarra et al., 2008). The particular and complex nature of the volatile substances, its variability on the time, the strong influence from atmospheric conditions and the subjectivity of smell perception are elements which delayed their regulamentations (Zarra et al., 2008). Nowadays the available techniques for ambient odour measurements are classifiable in analytical, sensorial and sensor-instrumental (Belgiorno et al., 2012; Zarra et al., 2014). Analytical measurements allow the characterization of odours in terms of chemical composition (GC-MS, colorimetric methods). Sensory measurements, such as dynamic olfactometry standardized by EN13725:2003, provide for using human nose as sensor, with a view to defining and measuring the effects of the odours on a panel of qualified examiners. Sensor-instrumental techniques allows defining information about the chemical composition and the smell propriety of the investigated odour. Between this last instrumental class, the Electronic Nose (e.nose) appears as the one with the most suitable potential. In fact, the use of the e.nose technology allows having a continuous and real time (or “near real time”) monitoring of odours (Romain et al., 2010; Munoz et al., 2010). According to Gardner and Bartlett (1994) an e.nose system consists of an array of non-specific gas sensors, a signal collecting unit and a pattern recognition software. The „heart‟ of the E.Nose technologies is their measurement chamber with inside the sensor array, designed to detect and discriminate complex odour mixtures (Rock et al., 2008). Different numbers of sensors are used to create the characteristic response called “fingerprint”. Moreover different types of sensors are available in the commercially market using a range of materials, including metal oxides, conducting polymers, surface acoustic wave devices and catalytic metals (Rock et al., 2008; Munoz et al., 2010). The principal steps implemented to use e.nose technologies consists of an initial training phase and a subsequent on site measurement application. The training of the e.nose represents the most important phase of the whole process (Romain et al., 2010; Giuliani et al., 2012). The goal of the training phase is to create the site-specific „odour measurement model‟ (OMM) that are robust, repeatable and reliable. For its definition, different statistical techniques can be used, in order to process the sampled data. A qualitative and a quantitative OMM were usually elaborated. The qualitative OMM have the objective to discriminate the different odour classes (e.g. odour sources) into a spatial domain. While the quantitative OMM have the finality to define the correlation equation between the electrical signal data,