| Abstract | Laser polishing (LP) is an emerging manufacturing process capable to address some of the significant limitations of traditional surface quality improvement technologies such as abrasive or chemical based polishing processes. By reconfiguring the topography of the outer surface, surface characteristics such as quality, aesthetics, wettability, friction, bio-fouling resistance, and others can be enhanced at a relatively low cost. However, numerous LP process parameters have to be fine-tuned to achieve the intended surface quality. This makes the selection of optimal process parameters time consuming and often unrepeatable. This study suggests that while both feed-forward and recurrent neural networks can be used to predict LP surface quality with a reasonable accuracy, the latter is characterized by a superior performance. |
|---|
