Artificial Neural Network-Based Surface Reconstruction Model of Wire-Arc Additively Manufactured Surfaces using Discrete Cosine Transfer

Abstract

Abstract The unique arc-based additive manufacturing process boasts exceptionally high deposition rates, but it comes with a trade-off – the generation of highly irregular surfaces that demand extensive post-machining efforts. These post-processing activities not only consume critical raw materials but also substantial energy resources. To mitigate these challenges and enhance process efficiency, it is crucial to develop predictive capabilities for surface topography based on key process parameters like wire feed speed and travel speed. In this study, a surface topography model is meticulously developed to reconstruct surfaces, utilizing wire feed speed, travel speed, and inter-pass temperature as input parameters. The initial phase involves identifying pertinent surface features that collectively define the surface. Out of the features examined, eight representative attributes emerge, including spatial average roughness, spatial peak height, spatial maximum valley depth, spatial skewness, spatial kurtosis, maximum flatness, and waviness. The surface reconstruction model employs the discrete cosine transform (DCT), necessitating a minimum of 30 DCT parameters for accurate surface reconstruction. Additionally, an ANN model is established to predict DCT parameters based on wire feed speed, travel speed, and inter-pass temperature inputs. Validation using the 309L stainless steel test material demonstrates the model's impressive accuracy in predicting DCT parameters, enabling precise forecasts of overall surface topography and machining allowances. This model lays the foundation for simulation-based additive-subtractive process design by identifying optimal deposition conditions and corresponding machining parameters. Furthermore, it streamlines the integration of realistic surfaces into computational models for additive process simulations, offering significant potential for improving additive manufacturing processes.