The Study and Application of Gaussian Process Surrogate Model Based on Gradient Particle Swarm Optimized Hyper-Parameters

Abstract:

Constructing a high-fidelity surrogate model to optimize production process is often required to meet the requirement of manufacturing process programming ,one of the most popular techniques for the construction of such a surrogate model is that of Gaussian process surrogate model. In this paper, the development of a gradient particle swarm optimization is described, which aims to reduce the cost of this likelihood optimization by drawing on an efficient adjoint of the likelihood, and improve the precision of the model. A multimodal benchmark function was used to test, show that the tuning strategy can provide an accurate Gaussian process surrogate model. Based on LHS ,Gaussian process surrogate model (GP) and gradient particle swarm optimization algorithm (GPSO), a optimization model which is used for improving the quality of Al profile welding is built and utilized to obtain optimal multi-parameters of Al alloy profile extruding processes and moulds. Optimal solution is validated by experiment.

• Periodical: Advanced Materials Research (Volumes 634-638)
• Pages: 4011-4016
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.634-638.4011
• Online since: January 2013
• Authors: Wei Xia, Hong Chen Pan, Xiao Ping Liao
• Keywords: Aluminum Profile, Gaussian Process, Gradient Particle Swarm Optimization, Hyper-Parameters Tuning

References

[1] Rasmussen C. E. and Williams, C. Gaussian Processes for Machine Learning 2006, MIT Press: 25-51

[2] M. N. Gibbs McKay D. Efficient Implementation of Gaussian Processes 1997(Canvedish Laboratory, University of Cmabridge, Cambridge, England,)

[3] Mackay D.J.C. Introduction to Gaussian Processes 1998, Department of Physics,University of Cambridge, Cambridge ,England

[4] Morten Løvbjerg Thomas Kiel Rasmussen, Thiemo Krink. Hybrid Particle Swarm Optimiser with Breeding and Subpopulations [J]. Proceedings of the Third Genetic and Evolutionary Computation Conference (GECCO-2001). 2001: 607-614

[5] de Jong K A. An analysis of the behavior of a class of genetic adaptive systems 1975(Doctoral dissertation,University of Michigan)

[6] Donati L., Tomesani L. The prediction of seam welds quality in aluminum extrusion [J]. Journal of Materials Processing Technology. 2004, 153-154(0): 366-373

DOI: 10.1016/j.jmatprotec.2004.04.215

[7] Donati L., Tomesani L. The effect of die design on the production and seam weld quality of extruded aluminum profiles [J]. Journal of Materials Processing Technology. 2005, 164-165(0): 1025-1031

DOI: 10.1016/j.jmatprotec.2005.02.156

[8] Donati L., Tomesani L., Minak G. Characterization of seam weld quality in AA6082 extruded profiles [J]. Journal of Materials Processing Technology. 2007, 191(1-3): 127-131

DOI: 10.1016/j.jmatprotec.2007.03.073

[9] Xiaoping Liao, Xuelian Yan, Wei Xia,lynnetwwwdagriorglynnet. A fast optimal latin hypercube design for Gaussian process regression modeling. in Advanced Computational Intelligence (IWACI), 2010 Third International Workshop on. 2010, 474-479

DOI: 10.1109/iwaci.2010.5585160

[10] Wei Xia, Bin Luo, Xiao-ping Liao. "An enhanced optimization approach based on Gaussian process surrogate model for process control in injection molding" The International Journal of Advanced Manufacturing Technology. 2011: 1-14

DOI: 10.1007/s00170-011-3227-4