Modelling of direct metal laser sintering of EOS DM20 bronze using neural networks and genetic algorithms
Year: 2013
Type: conference-paper
Abstract: An attempt was made to predict the density and microhardness of a component produced by Laser Sintering of EOS DM20 Bronze material for a given set of process parameters. Neural networks were used for process-based-modelling, and results compared with a Taguchi analysis. Samples were produced using a powder-bed type ALM (Additive Layer Manufacturing)-system, with laser power, scan speed and hatch distance as the input parameters, with values equally spaced according to a factorial design of experiments. Optical Microscopy was used to measure cross-sectional porosity of samples; Micro-indentation to measure the corresponding Vickers’ hardness. Two different designs of neural networks were used - Counter Propagation (CPNN) and Feed-Forward Back-Propagation (BPNN) and their prediction capabilities were compared. For BPNN network, a Genetic Algorithm (GA) was later applied to enhance the prediction accuracy by altering its topology. Using neural network toolbox in MATLAB, BPNN was trained using 12 training algorithms. The most effective MATLAB training algorithm and the effect of GA-based optimization on the prediction capability of neural networks were both identified.
Authors: A. Singh, David E. Cooper, N. Blundell, Gregory John Gibbons, D. K. Pratihar
Language: English
Subfield: Mechanical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W1512553471