ICSPEA: Evolutionary Five-Axis Milling Path Optimisation

Mehnen, J.^{1, a}; Roy, R.^{1, b}; Kersting, P.^{2, c}; Wagner, T.^{2, d}

1)

Decision Engineering Centre, University Cranfield, Cranfield, United Kingdom

2)

Institut für Spanende Fertigung, Universität Dortmund, Baroper Str. 301, 44227 Dortmund

a) j.mehnen@cranfield.ac.uk; b) r.roy@cranfield.ac.uk; c) petra.kersting@isf.de; d) wagner@isf.de

Kurzfassung

ICSPEA is a novel multi-objective evolutionary algorithm which integrates aspects from the powerful variation operators of the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) and the well proven multi-objective Strength Pareto Evaluation Scheme of the SPEA2. The CMA-ES has already shown excellent performance on various kinds of complex single-objective problems. The evaluation scheme of the SPEA2 selects individuals with respect to their current position in the objective space using a scalar index in order to form proper Pareto front approximations. These indices can be used by the CMA-part of ICSPEA for learning and guiding the search towards better Pareto front approximations. The ICSPEA is applied to complex benchmark functions such as an extended n-dimensional Schaffer's function or Quagliarella's problem. The results show that the CMA operator allows ICSPEA to find the Pareto set of the generalised Schaffer's function faster than SPEA2. Furthermore, this concept is tested on the complex real-world application of the multi-objective optimization of five-axis milling NC-paths. An application of ICSPEA to the milling-path optimisation problem yielded efficient sets of five-axis NC-paths.

Schlüsselwörter

Multi-objective Optimisation, Mechanical Engineering, CMA-ES, SPEA2, Evolutionary Computing

Veröffentlichung

In: Proceedings of the 9th Annual Genetic and Evolutionary Computation Conference (GECCO-2007), 7.7.-11.7. 2007, London, Thierens, D.; et al (Hrsg.), ISBN 978-1-59593-697-4