Authors: Tamal Ghosh; Kristian Martinsen
Addresses: Department of Manufacturing and Civil Engineering, Norwegian University of Science and Technology, Teknologivegen 22, 2815 Gjøvik, Norway ' Department of Manufacturing and Civil Engineering, Norwegian University of Science and Technology, Teknologivegen 22, 2815 Gjøvik, Norway
Abstract: In textile industries, ring and rotor spinning activities are most crucial to the yarn production process, which consist of many parameters and responses. To optimise the said process, the optimal settings for the process parameters must be obtained. Multi-objective optimisation models for yarn production exist but these are product sensitive and expensive in terms of the computation and production cost. In this article, an iterative multi-objective deep-learning assisted optimiser is developed and a non-dominated search technique is employed to obtain the Pareto optimal sets of the process parameters, which could improve the yarn quality. Further a Kohonen's self-organising map (KSOM)-based model is introduced to investigate the correlations among the yarn production variables. The proposed method is successfully validated with case studies and shown to outperform the existing results.
Keywords: yarn production process optimisation; deep-learning model; artificial neural network; multi-objective optimisation; self organising map.
International Journal of Experimental Design and Process Optimisation, 2020 Vol.6 No.3, pp.234 - 252
Received: 28 May 2019
Accepted: 13 Apr 2020
Published online: 26 Feb 2021 *