Affiliations: JNTU Hyderabad & MIC College of Technology, Kanchikacherla, AP, India | Deparment of Mechanical Engineering, Matrusri Engineering College, Hyderabad, AP, India | Deparment of Mechanical Engineering, JNTU Hyderabad, Hyderabad, AP, India | Department of Mechanical Engineering, MIC College of Technology, Kanchikacherla, AP, India | Department of Mechanical Engineering, Chhatrapati Shivaji Institute of Technology, Durg, CG, India
Note: [] Corresponding author. Pandu R. Vundavilli, Department of Mechanical Engineering, MIC College of Technology, Kanchikacherla, AP, 521180, India. E-mail: [email protected]
Abstract: This paper introduces an intelligent system for the prediction of mechanical properties of silica-based resin bonded sand core system. The properties of sand cores, such as tensile strength, compression strength, shear strength and permeability depends upon various process parameters, namely percentage of resin, of hardener, number of strokes and curing time. In the present paper, Mamdani-based fuzzy logic (FL) approach is used to perform forward modeling, in which the outputs are expressed as the functions of input variables. Moreover, the performance of FL system depends on the knowledge base (KB), which consists of rule base and data base. Three different approaches have been developed in the present work. Manually constructed FL system is developed in the first approach, whereas in approach 2, genetic algorithm (GA) is used to optimize the data base and rule base of FL system developed in Approach 1. On the other hand in Approach 3, automatic evolution of rule is considered along with the use of GA to optimize data base and rule base. It is important to note that the developed fuzzy model uses triangular membership functions for fuzzification and centroid area method for de-fuzzification process. The developed FL system eliminates the need of extensive experimental work in selecting the most influential process parameters. The performances of all three approaches have been tested with the help of twenty test cases. It is to be noted that all three approaches, developed can be effectively used in foundry for making prediction. The results showed that the Approach 3 has outperformed the remaining two, in terms of prediction accuracy.
