Design and Development of a Novel Portable Optical System with a Shallow Neural Network to Characterise Liquid Turbidity

Abstract

Clean water is a very important resource to all living things. This paper describes the design and development of a novel optical system for water quality characterisation. The system utilises transmitted light as well as scattered light from a sample to determine the level of turbidity of the liquid sample. The developed system can also be used to perform quality control on food and beverage (F&B) products that are manufactured on a large scale. The system was tested on various liquids of different concentrations. This was then used to train a shallow neural network and tested on new samples to identify the level of contaminants. For a good liquid sample the system was able to provide a good estimate with an error of less than 2 % with the 180°detector and about 12 % based on the reflected light. In addition to this, it was also used to determine if the beverage sample has gone rancid. For such a sample both the detectors were able to indicate abnormality.