A deep learning solution for real-time quality assessment and control in additive manufacturing using point cloud data

Abstract

This work presents an in-situ quality assessment and improvement technique using point cloud and AI for data processing and smart decision making in Additive Manufacturing (AM) fabrication to improve the quality and accuracy of fabricated artifacts. The top surface point cloud containing top surface geometry and quality information is pre-processed and passed to an improved deep Hybrid Convolutional Auto-Encoder decoder (HCAE) model used to statistically describe the artifact's quality. The HCAE's output is comprised of 9*9 segments, each including four channels with the segment's probability to contain one of four labels, Under-printed, Normally-printed, Over-printed, or Empty region. This data structure plays a significant role in command generation for fabrication process optimization. The HCAE's accuracy and repeatability were measured by a multi-label multi-output metric developed in this study. The HCAE's results are used to perform a real-time process adjustment by manipulating the future layer's fabrication through the G-code modification. By adjusting the machine's print speed and feed-rate, the controller exploits the subsequent layer's deposition, grid-by-grid. The algorithm is then tested with two defective process plans: severe under-extrusion and over-extrusion conditions. Both test artifacts' quality advanced significantly and converged to an acceptable state by four iterations.

On-Machine LTS Integration for Layer-Wise Surface Quality Characterization in MEX/P

“…Although particularly popular in metal AM [21][22][23], layer-wise monitoring has also been applied in MEX/P, primarily through three different sensing technologies: thermal sensing [24], 2D vision [25,26], and 3D vision [16,23,[27][28][29][30][31][32][33][34][35][36][37][38][39][40]. Monitoring the 3D layer topography offers distinct advantages over thermal and two-dimensional sensing, as it allows for direct measurement of layer features, like layer height [27,28], or in-plane defects [16,23,30,31].…”

“…The analysis of previous research efforts in LTS-based layer-wise monitoring of MEX/P part quality leads to the definition of several key aspects, like the integration strategy, the vertical repeatability, or the objective quality characteristic. The summary in Table 1 shows how most authors decided to integrate commercial LTS models [16,[28][29][30]35,37,38], although there are also examples of built-in-house sensor arrangements [27,36]. The layer-wise inspection could take place using pure on-machine solutions, where the LTS has been fully integrated into the manufacturing equipment [16,27,30,31,36] or has been attached to actuated [28,37,38] or fixed [29,35] off-machine arrangements.…”

“…The summary in Table 1 shows how most authors decided to integrate commercial LTS models [16,[28][29][30]35,37,38], although there are also examples of built-in-house sensor arrangements [27,36]. The layer-wise inspection could take place using pure on-machine solutions, where the LTS has been fully integrated into the manufacturing equipment [16,27,30,31,36] or has been attached to actuated [28,37,38] or fixed [29,35] off-machine arrangements. The scanning strategy and the relative movement between part and LTS are conditioned by the type of integration used.…”

“…Thus, solutions with low vertical resolution are applied to part height monitoring [27] or coarse defects [30,35]. LTS with intermediate performance are more versatile and have been used to evaluate the underfill and overfill issues [28,35,38], layer height deviations [34,35,37], warpage [37] or even the accuracy of the deposition trajectory [16]. On the other hand, the characterization of surface texture requires a high-resolution LTS and has been mostly beyond the scope of research efforts in this field, although its usefulness has been explored in off-machine arrangements [29].…”

Multi‐source data‐driven approach for prediction of melt density during polymer compounding

“…However, with the rise of deep learning, more and more researchers are applying this technology to the field of polymer processing detection. 16,17 The emergence of deep learning has revolutionized the landscape of polymer processing detection, enabling researchers to harness the capabilities of NNs and advanced deep learning techniques. By leveraging the power of NNs and other deep learning techniques, researchers across different fields can more effectively analyze large datasets and extract meaningful insights from complex systems.…”