Online Detection and Prediction of Fused Deposition Modelled Parts Using Artificial Intelligence

General

Multimaterial, multifunctional designs with complex geometric features with additive manufacturing (AM) have become possible within the last decade. Fused Deposition Modelling (FDM) is one of the most widely used due to its ease of use and low cost. FDM divides CAD models into thin layers of 2D patterns, which are then represented with extruded polymer rasters to create the final product. However, interlayer imperfections still cause delamination and warping, necessitating reprinting and wasting material. A weak adhesion between two layers can lead to delamination, and residual thermal strain from printing can cause warping. Warping can also cause delamination. Imperfections' appearance can be greatly influenced by printing settings such as print parameters, first-layer calibration, and model geometry. Recent advances in the application of artificial intelligence (AI) and machine learning to materials science and engineering problems have enabled researchers to use machine-learning algorithms to classify and predict various printing defections, such as blobs, warps, and delamination. The implementation of a new slicing mechanism to divide prints into spatially locked bricks in order to reduce warping is another intriguing development in the field. Contrary to previous research, which was unable to identify or predict interlayer problems before they spread throughout the entire construction process, this new study does so. This chapter discusses the use of a computer vision and strain gauge model to detect and predict interlayer imperfections in 3D printed parts, such as delamination and warping.

Delamination conditions can be classified and detected using camera-based images and deep learning algorithms. The researchers have also developed a new method of measuring and forecasting future warping based on strain gauge measurements. The first type of interlayer flaw to consider is delamination. Delamination is primarily the result of an incorrect gap between the current nozzle height and the print, which leads to a weak bond between the print's layers. Thus, the literature shows that the first step in resolving this issue is to properly calibrate the nozzle offset value (see Figure 1). Because manual calibration relies so heavily on the user's ability to see the shape of the extruded polymer at the nozzle tip with their naked eye, we've created a system that uses deep learning to mimic manual calibration. An in-house designed and printed camera mount holds a Logitech USB camera to the printing nozzle's left side, allowing the team to monitor the print process in real time. Vibrations caused by printing have been reduced thanks to the mount's cantilevered component being reinforced on the back. As a result of this, the filming angle is nearly horizontal, and the camera's plastic shell has been removed to make room for it in the printing space. Because the offset nozzle height of the first layer remains within the range of 0.1–0.2 mm, a near horizontal view is preferable to a far horizontal view. Now the distance between the raster's current position is more precisely monitored.

Defects in 3D Printing

Sensors and feedback controls are only available for the nozzle and build platform temperatures on the vast majority of Fused Filament Fabrication (FFF) 3D printers. Some operational prerequisites, such as filament detection, may require functional check sensors to be present. Commercial (FFF) 3D printers usually don't monitor the printing process or check to see if it's going as smoothly as anticipated. In the event of a defect or even a critical failure, such as the print object losing adhesion to bed, the printing process will continue unless the operator intervenes. This wastes material, power and equipment operation time, and it may also lead to malfunctions in the printing parts that are being used (e.g., nozzle clogging). The result is that print progress must be checked by process operators on a regular basis.