Abstract
Direct Metal Laser Sintering (DMLS) is an additive manufacturing process where metal parts are created layer by layer. Mechanical properties of the final product can vary significantly based on processing parameters. In traditional processes, such effects of processing parameters on mechanical properties are well-established. However, additive manufacturing methods are relatively new, which means there is less consensus, if at all, on how processing parameters affect mechanical properties of the final product. This study is a preliminary effort toward understanding the effects of processing parameters on mechanical properties of the metal. Processing parameters studied were the fabrication direction and temperature. Mechanical properties that were studied were the yield and tensile strength of the built material. 15-5PH stainless steel parts were DMLS fabricated with varying temperatures and directions for this purpose and their mechanical properties were measured. Then, a statistical approach was followed in order to generate a probabilistic prediction model. In this approach, Gibbs sampling was used to randomly sample from population of coefficients, Metropolis algorithm was used for decision-making purposes based on performance of different coefficient sets, and an empirical model was hypothesized. Then, the model was trained using a training dataset, and the cloud of coefficient sets for the hypothesized equation were obtained. Using these coefficient sets, the probable normal distribution of other test conditions was predicted and verified using testing data. It was shown that all measurements were well within the confidence interval of predictions, with a maximum difference of 4% between mean predictions and measurements. It was also observed that with a coefficient of variation smaller than 18%, spread of predictions was low enough to suggest that predictions were precise as well as their accuracy.
