Wang, Zimo (2020-08). Smart Sensing in Advanced Manufacturing Processes: Statistical Modeling and Implementations for Quality Assurance and Automation. Doctoral Dissertation.
With recent breakthroughs in sensing technology, data informatics and computer networks, smart manufacturing with intertwined advanced computation, communication and control techniques promotes the transformation of conventional discrete manufacturing processes into the new paradigm of cyber-physical manufacturing systems. The cybermanufacturing systems should be predictive and instantly responsive to incident prevention for quality assurance. Thus, providing viable in-process monitoring approaches for real-time quality assurance is one essential research topic in cybermanufacturing system to allow a closed-loop control of the processes, ensure the quality of products, and consequently improve the whole shop floor efficiency. However, thus far, such in-process monitoring tools are still underdeveloped on the following counts: o For precision/ultraprecision machining processes, most sensor-based change detection approaches are reticent to the anomalies since they largely root in the stationary assumption whilst the underlying dynamics under precision machining processes exhibit intermittent patterns. Therefore, existing approaches are feeble to detect subtle variations which are detrimental to the process; o For shaping processes that realize complicated geometries, currently there is no viable tool to allow a noncontact monitoring on surface morphology evolution that measures critical dimensioning criteria in real time. o For precision machining processes, we aim to present advanced smart sensing approaches towards characterizations of the process, specifically, microdynamics reflecting the fundamental cutting mechanisms as well as variations of microstructure of the material surfaces. To address these gaps, this dissertation achieves the following contributions: o For precision and ultraprecision machining processes, an in-situ anomaly detection approach is provided which allows instant prevention from surface deterioration. The method could be applied to various (ultra)precision processes of which most underlying systems are unknow and always exhibit intermittency. Extensive experimental studies suggest that the developed model can detect in-situ anomalies of the underlying dynamic intermittency; o For shaping processes that require noncontact in-process monitoring, a vision-based monitoring approach is presented which rapidly measures the geometric features during forming process on sheet-based workpieces. Investigations into laser origami sheet forming processes suggest that the presented approach can provide precise geometric measurements as feedback in real time for the control loop of the sheeting forming processes in cybermanufacturing systems. o As for smart sensing for precision machining, an advanced in-process sensing/ monitoring approach [including implementations of Acoustic Emission (AE) sensor, the associated data acquisition system and developed advanced machine/deep learning methods] is introduced to connect the AE characteristics to microdynamics of the precision machining of natural fiber reinforced composites. The presented smart sensing framework shows potentials towards real-time estimations/predictions of microdynamics of the machining processes using AE features.