Engineering-driven data analytics for In situ process monitoring of nanomanufacturing
MetadataShow full item record
Carbon Nanotubes (CNTs) buckypaper is a multifunctional platform material with superior mechanical and electrical characteristics. One of the critical roadblocks to scale-up production of high-quality buckypaper is the in situ process monitoring. This dissertation focuses on developing systematic methodologies for effective data analytics and process monitoring of nanomanufacturing. A novel generalized wavelet shrinkage (GWS) method was proposed to realize data denoising and signal enhancement for in-line Raman spectroscopy. A penalized mixed-effects decomposition (PMD) was developed to perform data decomposition and solve the multichannel profile detection problem in nanomanufacturing. We also proposed a novel tensor mixed-effects (TME) model to do high-dimensional data analytics for massive Raman mapping data with complex structure. In addition, different algorithms for parameter estimation were developed for these three approaches. Furthermore, by using numerical simulation and case study, we evaluated the performance of the proposed approaches. The GWS can significantly increase the signal-to-noise ratio as well as improve the accuracy and efficiency of Raman spectroscopy. The PMD lays a solid foundation for monitoring fabrication consistency, uniformity, and defect information, simultaneously. The TME provides us the capability to analyze massive high-dimensional data with mixed effects, and explore the complex correlations. It can be used to quantify the degree of alignment of CNTs buckypaper.