Abstract:
Rapid media filtration is used to treat most surface water sources for municipal
drinking water supply. This thesis presents results of an investigation into how
media filters in Scottish Water, built decades ago to different standards, can meet
increasingly stringent water quality requirements.
A new method for efficient assessment of filter performance and diagnosis of
issues using online data were developed. A recursive partitioning algorithm
applied to operationally relevant predictor variables was shown to efficiently and
effectively characterise the conditions associated with elevated turbidity over an
extended period of operation. Tree models can then be used to communicate a
diagnosis in operational terms to aid the efficient management of individual
pathogen barriers in a multi-barrier system.
Robust rapid filtration requires effective coagulation. An investigation was
conducted at a water treatment works (WTW) to understand the influence of zeta
potential. The effective zeta potential window was modelled & observed to
change with conditions. The online measurement of zeta potential was shown to
be useful for process optimisation by providing a quantitative measurement with
a mechanistic basis for coagulation conditions. This provides advantages over jar
testing which may poorly represent the system under investigation.
Pilot and full-scale trials of an alternative expanded aluminosilicate media show
that the additional bed expansion achieved by replacing sand with the lower
density material can, at comparatively low cost, improve the performance of rapid
gravity filters with a common design limitation. The effective application of new
finer grade expanded aluminosilicate was shown at pilot and full scale.
This thesis presents a suite of efficient solutions to ensure aging 20th century
filters robustly meet challenges in 21st century. This is achieved by improving
understanding of how and why specific constraints limit the performance of
existing RGFs and developing strategic solutions to overcome common
limitations to process performance.