Optimized deep neural network strategy for best parametric selection in fused deposition modelling

Fused deposition modeling (FDM) is a model of additive manufacturing (AM) which uses layer by layer-based methodology to fabricate a component. In the current digital manufacturing era, FDM process is widely used as it can construct intricate and complex part geometries in short time, its simplicity and economical behavior as compared to conventional manufacturing. Despite of such advantages, literature argued various machine learning approaches adopted to increase the performance of FDM addressing the issues of irregularities in part properties, accuracy, and reliability due to challenging task of best parametric selection. In this context, the present study proposed a deep neural network strategy to predict the best parametric combination with optimized mechanical properties (tensile and compressive strength) of printed parts. In the present research, the design variables as nozzle diameter, width of print line and layer thickness, print speed are considered as input parameters with their levels values that are trained to the proposed system. Adhering to ASTM standards with predefined dimensions total 256 experiments have been carried for each output, in which 204 result data used for training and 52 for testing the model using PYTHON programming language. Subsequently, the proposed model has gained the accuracy of 88.46% and root mean square value as 0.3396 is validated by relating the performance with existing models. Hence, the efficient outcomes of the developed model have been verified by gaining the best combination of process parameters and Taguchi analysis interpreted their influence on the tensile and compressive strength of FDM printed parts.