DETECTING PART ANOMALIES INDUCED BY CYBER ATTACKS ON A POWDER BED FUSION ADDITIVE MANUFACTURING SYSTEM

Abstract

Additive manufacturing systems are highly vulnerable to cyber attacks that sabotage parts and print environments during the designing, slicing and printing steps of the process chains. Due to the complex cyber-physical nature of additive manufacturing systems, cyber attacks are difficult to detect and mitigate, and impossible to eliminate entirely. Therefore, it is imperative to develop rapid and reliable non-destructive testing methods for detecting anomalies in printed parts.

This chapter describes a novel anomaly detection method developed for a selective laser sintering type of powder bed fusion system. The method does not engage computing-intensive machine learning to detect anomalies, relying instead on three side channels, print bed movement, laser firing time and print chamber temperature, that underlie the physics of selective laser sintering. The side channels provide adequate detection coverage while reducing the sensor requirements; they are also robust to noise, which enhances the detection of printed part anomalies. Experimental results demonstrate the efficacy of the anomaly detection method under attacks that target the mechanical properties of printed parts. The cost of the sensors and peripheral devices is minimal and anomaly detection for each test part requires less than three seconds.