Study and Development of Nature Inspired Algorithms for Intelligent Manufacturing Systems

Abstract

Quality improvement and cost reduction are two important objectives of intelligent manufacturing systems (IMS). Selection of optimal process parameters of any manufacturing process can improve the quality and reduce the cost. In this thesis effort is made to develop two variants of Artificial Bee Colony (ABC) algorithm to solve engineering design and manufacturing process optimization problems. The salient features of the research are: newline A new Artificial Bee Colony - Differential Evolution (ABC-DE) algorithm that embeds the Differential Evolution (DE) operators into the standard ABC algorithm is developed. newline A new Quantum Inspired Artificial Bee Colony (QIABC) algorithm that integrates quantum rotation operator into the standard ABC algorithm is developed. newline The ABC-DE and QIABC algorithms are tested on twenty-one test functions and the algorithms exhibited either improved or comparable performance over other state of the art algorithms in terms of standard performance metrics such as best, worst, mean, number of fitness evaluations and standard deviation. newline The ABC-DE and QIABC algorithms are applied to solve engineering optimization problems such as design of welded beam, pressure vessel, spring, gear train and three bar truss, and manufacturing process optimization of straight cylindrical turning, tangential turn-milling, CNC turning, Multi-pass turning, tool wear model of turning and ceramic grinding. The ABC-DE and QIABC algorithms exhibited improved or comparable performance over the state of the art algorithms in terms of standard performance metrics. newline Friction Stir Welding (FSW) process models of ultimate tensile strength, yield strength and % elongation are developed with response surface methodology (RSM) and adaptive neuro-fuzzy inference system (ANFIS) methods using the experimental data for joining dissimilar aluminum alloys AA 6061 and AA 6351. FSW process was optimized by ABC, ABC-DE and QIABC algorithms and the results are experimentally validated. newline newline