DIAGNOSIS OF CONCRETE CRACKING USING AN EXPERT SYSTEM

Abstract

An “Expert System" is a computer program which endeavours to model the knowledge and reasoning process of a human expert in a well defined area. Significant progress has been made in the development of expert systems for diagnosis in human medicine, and the use of expert systems in building defects diagnosis appears to be a promising line of development. In Singapore, concrete cracking has been a common but complex problem faced by building maintainers. The causes of cracking are many and varied. A careful diagnosis and analysis is crucial in determining a suitable repair technique, otherwise, the repair may only be temporary and uneconomical. Due to the scarcity of leading experts in this area, there is a strong need to capture the wide body of expertise needed to provide remedial solutions to concrete cracks on a reliable, continuing and widespread basis. This thesis explores the expert system technology to capture the expert knowledge in concrete cracking to provide methods for the diagnosis of cracks and advice on their repair techniques in a Personal Computer (PC) environment. The author examined the concepts and the problem solving strategy for the diagnosis of concrete cracking with the aim of developing a task model for acquisition and representation of knowledge in this domain. Two-step diagnostic goals, namely the identification of the cause of cracking and the selection of a remedial action, were defined. Using the task model, the acquired domain knowledge was analysed and systematically structured, and by carefully considering a number of parameters proposed by the author, a total of 8 causes of cracks and 23 remedial actions were proposed. The nature of diagnosis is goal-driven and backchaining data and this prompted Xi-Plus, a rule-based expert system programming tool to be chosen as the implementation platform for the system. The research reported in this thesis demonstrates the successful application of expert system techniques in providing diagnosis and remedial advice of concrete cracking in the PC environment. The prototype system XCRON runs interactively with the help of built-in explanation facilities, requires no external guide or manual, and is easily maintainable with the modular representation of knowledge in the form of rules. Overall, XCRON provides a standardised approach to diagnosing cracks in concrete, an area which is fairly unstructured and requires expertise. However, due to time constraints, the study was limited to selecting a remedial action for cracks found in beams, columns, reinforced-concrete walls, and the top and bottom faces of floor slabs that are common in a tropical climate like Singapore. This limitation, together with several other recommendations, are put forth in this thesis to exploit the full potentials of applying expert system techniques to building defects diagnosis.