ORIGINAL ARTICLE Ant colony optimization for group technology applications Anil Kumar Agrawal & Prabhas Bhardwaj & Vivek Srivastava Received: 25 November 2009 / Accepted: 29 November 2010 / Published online: 21 December 2010 # Springer-Verlag London Limited 2010 Abstract The problem of grouping of parts into part-families and that of machines into machine-cells has attracted the attention of many researchers particularly for medium variety of parts with medium production volume requirement, which traditionally required their production in batches for achieving economics of production. This kind of grouping, consequently offering benefits of mass production, was aimed to have independent cells processing ideally almost different sets of part-types. For this purpose, a number of approaches are available from various kinds of heuristics to mathematical programming formulations. Evolutionary methods such as neural network, genetic algorithm, and simulated anneal- ing have also been tried and have been found to provide better grouping solutions with much less computational complexity. In the present paper, ant colony optimization approach with number of newer strategies, incorporating more generalised framework of ants’ behaviour, has been applied to the parts and machines grouping problems taken from the literature. The results obtained from their application were found to be encouraging and thus establish the usefulness of the proposed approaches. Average performance of Tabu search with multiple ants was found to be the best and thus the parameter values for this approach were also determined using design of experiments methodology. Keywords Group technology . Ant colony optimization . Evolutionary methods . Design of experiments Abbreviations GT Group technology PMIM Part-Machine Incidence Matrix EE Exceptional elements CFP Cell formation problem ACO Ant colony optimization ACO-E ACO with elitist ant ACO-MA Basic multiple ant approach ACO-R ACO with ranking ACO-TS ACO with Tabu search ACO-SE ACO with shared experience 1 Introduction In group technology (GT), major challenge faced by the researchers is in formation of part-family and machine group and hence a cell. GT works on the concept of similarity between various types of parts. This means all similar parts are required to be processed in a cell utilising the same assorted group of machines. To achieve this objective, information used is mostly process-plan or route- plan data. This information is generally summarised in the form of part-machine incidence matrix (PMIM) used for finding the cells of parts and machines. These cells appear in the diagonalized sub-matrices in the PMIM. In the ideal condition, these cells must be totally independent to each other in order to get the benefits as mentioned above. This A. K. Agrawal : P. Bhardwaj (*) Mechanical Engineering Department, Institute of Technology, Banaras Hindu University, Varanasi 221005, India e-mail: prabhasbhardwaj@yahoo.co.in e-mail: pbhardwaj.mec@itbhu.ac.in A. K. Agrawal e-mail: bhu.anil@gmail.com V. Srivastava Senior Software Engineer, Ubona Technologies, Bangalore, India e-mail: viveksri15@gmail.com Int J Adv Manuf Technol (2011) 55:783–795 DOI 10.1007/s00170-010-3097-1