Clean water is a finite resource, and the quality of such is best monitored by colorimetric in-situ sensors, which allow frequent, non-labour intensive sampling, and are low-cost and simple to manufacture. There are multiple types of sensors that exist in the literature, however, many are cost-prohibitive for wide deployment, or the literature does not not fully elaborate on their operation. The aim of this research was to extend the lifetime and improve the performance of a colorimetric in-situ sensor, Aquamonitrix colorimetric sensor, that was produced by T.E. Laboratories, in addition to characterising sensor behaviour. Its operation was focused on the Griess reaction, in which a vivid azo dye is produced in the presence of nitrite, that can be linearly calibrated to the absorbance by the dye from a photodectector placed at the opposite end of a microfluidic detector channel
to a monochromatic light source. Using multiobjective optimisation on a numerical model of a Y-junction micromixer, it was found that both sensitivity could be increased and reagent could be conserved, by limiting the proportion of reagent used during testing to 5% to 7.5% of testing solution, as opposed to the 50% originally used by the system.
The conservation of the reagent allowed for an increased sensor deployment lifetime
of up to tenfold. To better understand how the parameters of analyte concentration, reagent
proportion of test solution, and mean flow velocity of the solution affect sensor
output, both mechanistic and data-based modelling of the continuous and stopped flow
stages of the sensor were undertaken. Third-order and second-order models were identified
for the continuous and stopped flow data respectively. The second-order model is
analagous to the two-step Griess reaction, of which there is a first, faster step. Further
characterisation of the zeroes, poles and transfer function coefficients of the third order
models showed that parameterisation was possible and, using principal component
analysis, reduction of parameters. Other testing on the effects of order of cycles, turbidity
and heavy metals was also conducted to measure their impact on sensor output.
Carryover between sensor cycles was found to be the most interfering factor on sensor
output, due to the microfluidic connector components, which was eliminated after eight
cycles. Overall, the performance and efficiency of the existing sensor was improved,
iii
and the methodologies in this dissertation can be used for other continuous-flow colorimetric
sensors and reactions, or even other microreactor applications, such as in green
chemistry.