In-Situ Quality Assessment and Control in Additive Manufacturing Using Laser Surface Profilometer and Deep Learning Techniques

Abstract

Abstract Additive Manufacturing (AM) has become one of the most popular manufacturing techniques in various fields. Their layer-by-layer printing process allows easier fabrication of complex geometries. However, the quality and accuracy of fabricated artifacts in these techniques have low repeatability. In the era of Industry 4.0 by using emerging sensory and data processing capabilities such as Laser Surface Profilometer (LSP) and Deep Learning (DL), it is possible to improve the repeatability and quality of AM processes. This work presents an in-situ quality assessment and improvement using LSP for data acquisition and DL for data processing and decision making. The utilized LSP module generates a point cloud dataset containing information about the top surface geometry and quality. Once the point cloud data is pre-processed, an improved deep Hybrid Convolutional Auto-Encoder decoder (HCAE) model is used to perform the artifact's quality measurement and statistical representation. The HCAE model's statistical representation is comprised of 9*9 segments, each including four channels with the segment's probability to contain one of four labels, 1) Under-printed region, 2) Normally printed region, 3) Over-printed region, 4) Empty region. This data structure plays a significant role in determining the commands needed to optimize the fabrication process. The implemented HCAE model's accuracy and repeatability were measured by a multi-label multi-output metric developed in this study. The assessments made by HCAE are then used to perform an in-situ process adjustment by manipulating the future layer's fabrication through the G-code modification. By adjusting the machine's print speed and feedrate, the control algorithm exploits the subsequent layer deposition, segment by segment. The algorithm is then tested with two settings that can cause defective parts: severe under-extrusion and over-extrusion conditions. Both test artifacts' quality advanced significantly and converged to an acceptable state by four iterations.