Filament extrusion state monitoring in fused filament fabrication using Bhattacharyya distance and K-nearest neighbor

Abstract

Fused filament fabrication (FFF) is one of the most popular techniques of additive manufacturing. However, product quality issues still limit the further application of FFF technology. Filament extrusion state has a great influence on the quality of FFF fabricated products, since both under-extrusion and over-extrusion can lead to the deterioration of product quality. Therefore, monitoring the filament extrusion states is vital and essential. This paper aims to monitor the filament extrusion state by acoustic emission (AE). To achieve this goal, experiments are conducted on a desktop FFF machine, where the states of under-extrusion and over-extrusion are induced by different extrusion speeds. Original AE signals are collected during the experiments. Confronted with the challenge posed by the susceptibility of AE signals to noise during the complex extrusion process and different conditions, one calculates the statistical distribution of the features defined on the raw AE signals, without the need for noise reduction steps. The k-nearest neighbor algorithm is then adopted to identify the different extrusion states, where the Bhattacharyya distance is employed to measure the distances or similarities of the calculated distributions. The findings demonstrate the successful identification of various extrusion states induced by different extrusion speeds through the presented method. The outcomes of this study pave the way for the development of an affordable in-situ FFF monitoring system with comprehensive capabilities.