Saideep Nannapaneni, Sankaran Mahadevan, Abhishek Dubey,
Recent technological advancements in computing, sensing and communication have led to the development of cyber-physical manufacturing processes, where a computing system monitors the manufacturing process performance in real-time by analyzing sensor data, and implements the necessary control to improve the quality of the manufactured product. In reality, the product quality may be affected by several uncertainty sources that may arise from the computing system, manufacturing process, and data collection such as sensor uncertainty, computing resource uncertainty, control input uncertainty, natural variability and modeling errors. Due to the continuous interactions between the computing system and the manufacturing process, these uncertainty sources may aggregate and compound over time, resulting in undesirable product quality variations. Therefore, characterization of the various uncertainty sources and their impact on the product quality needs to be investigated to improve the process efficiency and product quality. Towards this objective, this paper develops a framework for aggregating multiple uncertainty sources using a two-level dynamic Bayesian network, where the higher level captures sensor uncertainty, control input uncertainty, variability and modeling errors in the manufacturing process while the lower level captures the uncertainty related to the computing system. In high-dimensional manufacturing, as the number of process parameters increase, online analysis for control can be computationally expensive. To alleviate the high computational expense, variance-based global sensitivity analysis is used for dimension reduction. The proposed methodologies of online monitoring and control under uncertainty, and dimension reduction, are illustrated for a cyber-physical turning process.
Journal of Intelligent Manufacturing
cybermanufacturing, cyber-physical, process monitoring, control, uncertainty, Bayesian network