Real-Time Defect Detection in 3D Printing and Analysis in Additive Manufacturing Using Machine Learning

In order to identify architectural flaws in things made using additive manufacturing, this investigation suggested a method that utilized machine learning (ML) frameworks. The ML models are used to identify flaws in real manufacturing after being developed on artificial 3D point clouds containing faults. A substantial quantity of instruction effort and related expenses connected with several printing for every concept might be avoided by using synthesized clouds of three-dimensional points instead of data collected from experiments. This technique introduces a new notion termed “patch” to gather macroeconomic knowledge regarding neighboring points to facilitate neural network training and delivery, in addition to calculating the distance variations between distinct units across the source and the target cloud points. According to computational evaluations of forecasting outcomes on experiments with varying forms, the suggested approach worked better than the Z-difference technique that is currently in use in the available literature. aforementioned artificial intelligence techniques-Bagging of Trees, Gradient Boosting, Random Forest, K-nearest Neighbors, and Linear Supported Vector Machine-were tested in a range of scenarios with varying fault dimensions and data point volumes. In terms of reliability, it was discovered that the two most successful methods were The bagging procedure and Random Forest; 20 had been found to be the ideal patches number. With the use of an appropriate 3D data collecting structure, the suggested machine learning (ML)-based technique may be used for right away recognizing defects throughout additive manufacturing.