Construction, characterisation and analysis of results from a liquid fingerprinting system using ultrasound

About the Project

There is a need to be able to identify liquids in situ using small samples and with high reliability. The applications for such a system are very wide and range from validation and quality assurance in process industries, such as for drinks and fuels, for authenticating perfume compositions, analysis of water in a wide range of environments, as well as for process liquids such as oils or lubricants. It is also applicable to biological liquids, such as sweat, blood or urine, or those that are only available in small sample volumes, such as tears or cerebrospinal fluid. Our particular interests in the first instance relate to milk, soft drinks, alcoholic drinks and olive oils.

This project would involve further development and refinement of an earlier system, based on the use of ultrasound, to characterise liquids based on their mechanical and rheological properties as they form pendant drops. Such a system would involve physical construction of a working prototype, and it would require knowledge and proficiency in electronics as well as strong mechanical skills. Earlier work in this area [1] has proved that such a system can distinguish similar alcoholic drinks or dilutions of red wine [2,3], for example, or characterise specific aviation biofuels.

An important part of this project would be to analyse the large number of signals already recorded and which will also arise from measurements undertaken in this work using deep learning pattern recognition work, as already developed for use in other applications, such as classification of acoustic signals from knee joints [4]. These are likely to involve the use of Convolutional Neural Networks (CNNs), although other methods may also be employed.

The ideal candidate for this PhD position should have a strong background in applied physics, electronic engineering, or chemical analysis, and ideally proficiency in the application of deep learning methods, for instance in the Matlab environment. Commitment, enthusiasm and strong teamworking skills are essential, as well as the self-motivation to work independently as required.

Funding Notes

there is no funding for this project

References

1) A Wavelet, Fourier and PCA Data Analysis Pipeline: Application to Distinguishing Mixtures of Liquids M. Kokuer, F. Murtagh, N. D. McMillan, S. Riedel, B. O’Rourke, K. Beverly, A. T. Augousti, and J. Mason J. Chem. Info. Comp. Sci. 43 587-594 2003
2) A Wavelet, Fourier and PCA Data Analysis Pipeline: Application to Distinguishing Mixtures of Liquids M. Kökü er, F. Murtagh, N. D. McMillan, S. Riedel, B. O’Rourke, K. Beverly, A. T. Augousti, and J. Mason J. Chem. Info. Comp. Sci. 43 587-594.
3) Ultrasonic tensiographic measurements on liquid drops for liquid fingerprinting A. Augousti, J. Mason, H. Morgan and N. D. McMillan Sensors and Their Applications XII S. J. Prosser and E. Lewis (Eds) p297-302 IOP Publishing 2003
4) Deep Learning Successfully Classifies Acoustic Emission Data from Knee Joints I. Vatolik, C. Mbachu, G. Hunter, N. Swann, M. Everington and A. T. Augousti ISB23 1 August 2023 Fukuoka Japan