Prognostic fault prevention by segmented digital transformation of manufacturing process signals

Abstract

Faults during operation of a system can occur at any time. Contemporary fault diagnosis systems focus on identifying the problem. However, when faults occur, the system has already incurred losses, either as rejects or machine damage. In this paper, a new method that predicts future faults and subsequently prevents losses caused by the faults has been developed. The method makes use of a series of digital transformations of signals from the manufacturing process. An important finding of this research is to divide the continuous signal stream into data segments from which a parameter called sum standard deviation frequency (SSDF) can be extracted. SSDF reflects the change of inconspicuous conditions in the signal in different data segments over time. To predict how the signal stream will perform in the next minute, a new density peak clustering computational algorithm is developed to transform the SSDF database into “normal”, “marginal”, and “abnormal” records. Finally, a novel method “moving SSDF deep learning” (MSDL) concept has been proven to have a better prediction of future capability than four other prediction methods. The method has been applied to a 3D printing process in which belt tension affecting machine performance has been predicted 30 s ahead of the problem, giving sufficient time for the operator to adjust or stop the machine before a fault occurs.