KNOW LEDGE-BASED SYSTEM ON WATER= RESOURCE MANAGEMENT IN COASTAL WATERS K. W. Chau. MSc, M Phil, PhO, CEng. MICE, MlStrudE * ABSTRACT Computers which are used conventionally in numerical models for problem-solving and fast number-crunching are not user-friendly and lack knowledge transfer in model interpretation. Recent advances in artificial intelligence make it possible to systemise the experience and decision-making procedures o f human experts. This paper describes fhe development of a prototype knowledge-based system (an advanced technique of artificial intelligence) as a knowledge-transfer tool for water-resource planning and management in coastal wafers. The system can (a) run on a desktop computer and (b) act as a design aid, to assist in numerical modelling to simulate flow and/or wafer quality. Using the application in Hong Kong coasfal waters, it was verified that the prototype system had the capability to undertake fhe reasoning process to select the model and associated paramefers, equivalent to a human expert. Key words: Artificial intelligence; coastal waters; kno wledge-based system; knowledge transfer; water-resources management. * Associate Professol; Department of Civil and Sfructural Engineering, Hong Kong Polytechnic Universik Hunghorn, Kowloon, Hong Kong. INTRODUCTION Recent developments in artificial-intelligence technology make it possible, by encoding knowledge and reasoning, for a computer program to simulate human expertise in a narrowly defined domain during a problem-solving process. This type of program is designed to ~ assist in solving problems that require the skill and expertise of a I human, by the application of heuristic rules of thumb. It is generally referred to either as a knowledge-based system (KBS) or an expert system".2'. The capability of a KBS to make inferences from given relationships and to cope with uncertainty in complex situations has been used for a wide range of application^'^-'^'. However, literature on the incorporation of expert system technology into numerical modelling is limited. It is now generally accepted that KBS is particularly suitable for unstructured problems in a narrowly defined subject domain. Accordingly, it is believed that this application of information technology will have a significant effect on water-resource development. Also, a well-constructed KBS program can play an important role in knowledge transfer and the training of technologists and, perhaps (more importantly) in their subsequent professional ' activities. The role of a KBS as an expert involves (a) problem-solving whilst enriching the user's understanding of a problem, (b) assisting and supporting the user with relevant information, and (c) providing guidance. The system is intended for use by water-resource personnel who have little advanced knowledge about numerical modelling. i TOOLS IN WATER-RESOURCE MANAGEMENT In water-resource planning and management, it is necessary to be able to predict, with accuracy and efficiency, the flow and water quality in coastal waters. Tide prediction is important for the navigation and construction of engineering works. Also, the data on tidal currents can be used for evaluating pollutant transport and dispersal, or sediment deposition. The basic principle to describe flow and transport in a mathematical model uses the Navier-Stokes equationP along with necessary boundary conditions. Solving coupled partial-differential equations, owing to their non-linear nature, is a difficult mathematical THE JOURNAL I V18 N1 I MARCH 2004 25 problem. Numerical modelling is the best procedure for considering realistic features, e.g. the non-linearity, variation of roughness and bed shape of a fluvial system. However, the present technique for numerical simulation of flow and/or water quality is highly specialised, requiring detailed knowledge, and is not sufficiently user-friendly. The numerical technique can be based upon (i) the finite element method, (ii) the finite difference method, (iii) the boundary element method, or (iv) the Eulerian-Lagrangian method. The time-stepping algorithm can be implicit, explicit or characteristic-based, and the shape function can be first order, second order or a higher order. The modelling can be simplified into different spatial dimensions, i.e. 1-dimensional model, 2-dimensional depth-averaged model, 2-dimensional layered model, 3-dimensional model, etc. The analysis is usually effected through simplification and a modelling technique based upon the experience of specialists. However, the accuracy of the prediction depends upon (a) the accuracy of the open-boundary conditions, (b) the model parameters, and (c) the numerical scheme. It is a difficult task for a novice engineer to select an appropriate numerical model to suit varying factors under different conditions, e.g. water depth, water velocity and grid spacing. Fulfilling the above criteria in flow and/or water-quality simulations is a complex decision-making process. There are new demands on water-resource staff to apply the existing tools properly and intelligently, including providing appropriate input parameters. However, with the recent advent of artificial-intelligence technology, personnel can be trained to fulfil this requirement. A computer with sufficient knowledge storage and a user-friendly interface can be consulted as an expert; as a result, it is a suitable KBS application. Moreover, there are usually few specialists with a thorough understanding of numerical modelling of flow and/or water quality. The KBS enables staff to become acquainted with up-to-date simulation tools and fill the existing gaps between researchers and practitioners in the application of recent technology.