كليدواژه :
Sugeno , Dogleg severity , Fuzzy Logic , case-based reasoning
چكيده لاتين :
Application-oriented research in the area of case-based reasoning has moved mature research results into practical applications. This paper presents an overview of different applications of case-based reasoning (CBR) in petroleum engineering, with focus on the drilling process, based on a survey and comparative evaluation of different applications. The numbers of papers, research groups, and experimental systems are indicative of the importance, need, and growth of CBR in different industries. A clear growing trend has been seen in the oil and gas industry over the last 5–10 years. In this paper we present the evolving story of CBR applied to problems in drilling engineering. We show that drilling engineering is an application domain in which the systematic storage and situation-triggered reuse of past concrete experiences provide significant support to drilling personnel at various levels. Some CBR systems have been successfully deployed in operational settings. With increased understanding of the complexity of drilling operations and continuous development of CBR and combined methods, the future potential is significantly higher. In this paper, the potential effects from membership function features on Sugeno fuzzy inference system for dogleg severity prediction are studied. When membership functions are symmetrically distributed along input scale and output scale, and completed If-Then rules are evenly built, the combination of triangular input MFs with rectangular output MFs will produce fuzzy inference model with perfect linear performance. Through adjusting geometric features of rectangular membership functions, linear SISO model can be easily converted into various non-liner models. Finally, by increasing the quantity of fuzzy sets (MFs) for input variables, the linearity of input-output relation will be improved, and the controllability of non-linear fuzzy inference system will also be enhanced as well. All of these features of membership function are retested and verified on Mamdani fuzzy inference models through this paper
