Functional Boxplots for Outlier Detection in Additive Manufacturing

TL;DR

This paper introduces a novel functional boxplot method for real-time outlier detection in additive manufacturing, enhancing process monitoring by identifying defective products based on geometric deviations.

Contribution

The paper proposes a new functional boxplot approach for outlier detection in AM, enabling automatic, online identification of defective products using geometric data.

Findings

High detection accuracy for geometric outliers

Effective in real-time process monitoring

Reliable across various complex geometries

Abstract

Additive manufacturing (AM), also known as 3D printing, is one of the most promising digital manufacturing technologies, thanks to its potential to produce highly complex geometries rapidly. AM has been promoted from a prototyping methodology to a serial production platform for which precise process monitoring and control strategies to guarantee the accuracy of products are required. This need has motivated practitioners to focus on designing process monitoring tools to improve the accuracy of produced geometries. In line with the emerging interest, in the current investigation, a novel strategy is proposed which uses functional representation of in-plane contours to come up with statistical boxplots with the goal of detecting outlying AM products. The method can be used for process monitoring during AM production to automatically detect defective products in an online fashion. To ensure the considered method has an acceptable potential, different complex 3D geometries are considered and undergo different types of stochastic perturbations to collect data for outlier detection. The results of the conducted simulation are very promising and reveal the reliability of the proposed method for detecting products with statistically significant deformation.