As the scale of Industrial Internet of Things (IIoT) applications in discrete manufacturing has exponentially expanded over the past several years, there is a growing opportunity for employing general-purpose diagnostic algorithms that can robustly operate on a wide variety of machines (vertical mills, horizontal lathes, grinders, stamping machines, etc.) that are manufacturing a wide variety of parts (Swiss turned parts such as fasteners, connectors, gears, etc., medical devices, aircraft components, firearm components, etc.). In particular, the basic question of “Is the machine operating normally?” is both universal and often of critical importance.
In this paper, we explore the possibility of a machine- and part-agnostic algorithm for anomaly detection using CNC machine control data. Such an algorithm can become an important component in any number of diagnostic applications, including helping to predict or categorize machine failures, flagging manufacturing defects, and spotting corruption within the data stream itself.