Development of portable, application specific electronic nose for agriculture

Abstract

Research groups around the world are working to develop electronic nose systems that are able mimicking the functions and operations of the human nose. The instrument is used to identify and classify different types of odour or smell. The instrument will complement the existing odour assessment techniques; human sensory panels and Gas Chromatography Mass Spectrum (GC-MS) analysis which require long training time and detailed operating procedures. However most of the generic instruments are of laboratories type which are costly and may not operate efficiently for every possible application. The instruments’ broad non-specific sensor arrays’ will limit the detection capabilities. The existing portable instruments in the market are still lacking in reliability, data processing capabilities and quite costly. Therefore, the purpose of this research is to develop a portable Application Specific Electronic Nose (ASEN) to improve their capabilities. The developed instrument uses specific selected sensor arrays which were identified based on experiment and key volatile compounds of the target odorant. Humidity and temperature sensor are also being included in the instrument to measure the environmental condition. The instrument utilises multivariate statistical analysis (PCA, LDA and KNN) and Artificial Neural Network (ANN) as well as an embedded ANN classification algorithm for the data processing. This will increase the instrument’s capability while the portability will minimise the size, cost and operational complexity. A commercial instrument (Cyranose C320 from Smith Detection) is used to evaluate the performance of the instrument. The instrument was successfully developed, tested and calibrated odour samples with variable concentrations. The instrument provides a feasible alternative for non-destructive testing system for the odour samples. The results revealed that the developed instrument is able to identify, discriminate and classify the odour samples with an acceptable percentage of accuracy. This will contribute significantly to acquiring a new and alternative method of using the instrument for agriculture applications i.e., plant disease detection, food quality assurance and poultry farm malodour monitoring. The future works include the development of specific sensors for the application and simplified the training process i.e., performs on-line ANN training by the instrument itself.