Analysis of Non-Linear and Non-Stationary Data with Intrinsic Time-Scale Decomposition (ITD) and Gaussian Mixture Model: Application to Acoustic EmissionBased Process Monitoring
MetadataShow full item record
Real-time change detection and abnormal prediction play essential roles in the safety and quality assurance of manufacturing processes. This research tried to verify the efficiency of Dirichlet Process Gaussian State Machine (DPGSM) and Intrinsic Time-Scale Decomposition (ITD) applied on the Acoustic Emission (AE) cutting sensor signal — which is a kind of nonlinear, non-parametric, and non-stationary manufacturing process signal that is generated during the cutting process — to assess the characteristics of machining of a Natural Fiber Reinforced Plastic (NFRP) composite material. The research attempted to determine whether the combination method of DPGSM and ITD is eligible to distinguish the pattern change of the cutting process under process parameter settings. The research also employed Average Run Length 1 (ARL1) to justify the performance of the DPGSM and ITD methods, classify the cutting process based on the results of Control Chart plots, and relate the change points to the actual process through the use of videos and AE cutting signals gathered during an earlier experimented study of the cutting process of the NFRP material.
Guo, Ruiqi (2019). Analysis of Non-Linear and Non-Stationary Data with Intrinsic Time-Scale Decomposition (ITD) and Gaussian Mixture Model: Application to Acoustic EmissionBased Process Monitoring. Master's thesis, Texas A&M University. Available electronically from