Introduction systems. The technique generates a population of


Numerical models are
frequently used to simulate the flow and water quality problems. Usually,
selecting a suitable numerical model to solve a practical water quality problem
is a highly specialised task requiring detailed knowledge on the application
and limitation of models. Due to the complexity, there is an increasing demand
to integrate artificial intelligence (AI) with these mathematical models in
order to assist selection and manipulation.


The advancement in
artificial intelligence over the past few decades has made it possible to
integrate technologies into numerical modelling systems to remove constraints
produced by current numerical models which are not user friendly. Many users do
not have the adequate knowledge to get their input data for an algorithmic
model and evaluate their results. A problem in modelling is when a parameter is
change in a model, the results may vary. This may lead to inferior designs
causing a failure of the model. There are several algorithms and methods which
can be used, in this report, knowledge based systems, genetic algorithm,
artificial neural network and fuzzy inference system techniques are explored. Each
AI techniques has different algorithms thus has different applications in water
quality testing; each application was assessed to see which AI technique would
fit best for a specific task.


Knowledge based systems automate
the decision making and reasoning processes of human expects in solving
problems using logical reasoning. This technique uses a collection of general facts,
rules of thumb and casual models of the behaviour specific to the problem
domain. This can be used for the selection and manipulation of various
numerical models on water quality.

Genetic Algorithm uses
computational models based on natural selection, which drives biological
evolution in developing computer based problems solving systems. The technique
generates a population of points at each
iteration where the best point in the population approaches an optimal solution
1. Parameters of numerical models of water quality can be calibrated using
function optimization.

Artificial Neural Network
is a technique that uses data driven models approached with highly
interconnected processing elements. Artificial Neural Network uses an
information processing paradigm that is inspired by biological nervous systems
in simulating underlying relationships that are not fully understood. This can
be used for the determination of underlying physical/ biological relationships
that are not fully understood.

Fuzzy inference system is
a technique that uses fuzzy set theory to map inputs and outputs. When
objective or the constraints are vague, fuzzy inference systems use modelling
complex and imprecise systems. This can be used in water quality modelling when
the objectives and constraints are not well defined, for example, quantifying
the dissolved oxygen, oxidation-reduction potential, pH and temperature of

Future Directions

All the AI technique are able
to carry out specific task in relation to water quality testing, however, a versatile
technique is to combine all these techniques together. It is believed, with the
ever-heightening capability of AI technologies, that further development of
numerical modelling in this direction will be promising. More efficient AI
techniques will arise, addressing the key issues with the current AI techniques
and produce more user-friendly systems with a clearer knowledge representation.


This study has reviewed
the progress on the integration of AI into water quality modelling. The
integration of various AI techniques, including knowledge based systems, genetic
algorithms, artificial neural networks and fuzzy inference system, into
numerical modelling systems have been reviewed where it was found that these
techniques can contribute to the integrated model in different aspects and may
not be mutually exclusive to one another.