Brandon Haschick

Computer Science Honours 2017

The Development of a Quasi Real-Time Water Pollutant Detection System for the Internet of Things (IoT)

Supervisor: Mosiuoa Tsietsi

Abstract:

The Internet of Things (IoT) is a paradigm of computing that aims to enable machine to machine communication among small form-factor devices such as microcontrollers and microprocessors which are embedded in our daily environment. IoT is becoming prevalent in homes, vehicles and industrial applications, where massive IoT networks are expected to usher in the age of the ‘Smart City’. However, these networks, generally consisting of resource-constrained devices, come with limitations and special constraints that have to be considered when developing IoT systems.

One particularly important application of IoT is water pollution detection, particularly with issues such as water scarcity becoming a major issue in developing countries such as South Africa. In this domain, the process of voltammetry is used by physical chemists to determine the presence of analytes (and their concentrations) in a water solution and is often used at water treatment plants to detect pollutants. Currently, to perform their experiments, chemists must use large, bulky and expensive equipment in the form of a potentiostat. The purpose of this research was to replace such equipment by implementing a system that is significantly scaled down, cheaper, networked, with lower power consumption and added intelligence and pollutant classification features. This was done by comparing existing technologies and making appropriate design decisions based on these criteria. The technologies considered include the data format, database, wireless communication and messaging protocol, with the devices being compared being a variety of models of Arduinos and Raspberry Pis.

The system built implemented a point-to-point topology incorporating an Arduino Uno which sent readings via the ZigBee wireless protocol to a Raspberry Pi 2 Model B for processing and visualisation. The peaks that correspond to the expected potentials at which known pollutants peak were successfully identified to aid in the process of pollutant classification and were stored in a cloud database for historical analysis and processing. The limitation of the solution is that it can only be used as an aid to physical chemists who must use specialist domain knowledge to translate the potential readings to actual analytes and use further techniques to calculate the concentration of said analytes.