Process parameter optimization using a feed-forward neural network for direct metal laser sintering process

As one of the Rapid Prototyping (RP) processes, Direct Metal Laser Sintering (DMLS) method is used to build prototype parts by depositing and melting metal powders layer by layer. Using DMLS, metal powder can be melted directly to build functional prototypes. The resulting properties of interest to the users include the processing time, mechanical properties, and geometric accuracy. Some main process parameters could affect the results significantly. These process parameters, which involve the laser scan speed, laser power, hatch density, and layer thickness, can be determined by the operator before building the prototype parts. But the relationships between these parameters and resulting properties are complicated. In many cases, the effects of different parameters on the resulting properties contradict one another. A method based on the Feed-forward Neural Network(NN) is described in this paper for predicting the resulting properties of the laser-sintered metal parts. After continuous training by using the data pairs, this NN constructs a mapping relationship between the process parameters and resulting properties. The objective of this research is to obtain the statistical relationships of the selected process parameters and the achieved process results.