Asset integrity and predictive maintenance models require data for an accurate assessment of an asset’s condition. Historically this data is collected periodically in the field by technicians using hand-held units. The significant investment in inexpensive microelectromechanical (MEMS) sensors communicating wirelessly to the cloud is changing the way asset data is collected. Permanently installed MEMS-based sensing units will enable near-real-time data collection and reduce the safety exposure of technicians by eliminating the need to obtain field data manually.
The proposed project will develop a framework for assuring the validity of the data these MEMS-based units generate. Drawing on the work in Khatibisephr , Bayesian statistical methods will be used in conjunction with Soft Sensors (SS). SS usage is widespread in the process industry as reviewed by Kadlec et al . The use of a Bayesian framework for SS has been pioneered in the nuclear industry as demonstrated by Hines . While the research to date shows the suitability of the proposed method, there is a lack of research for the application of SS in the civil sphere. Furthermore, the impact of MEMS-based systems and their integration into validation frameworks will be examined. I intend to employ an interdisciplinary approach utilising my background in electronics and statistics. I will also seek to expand my knowledge of process control, machine learning and asset management to wholly describe the emerging field of permanently installed MEMS-based sensing units for condition monitoring and structural integrity.
 S. Khatibisepehr, “Bayesian Solutions to Multi-model Inferential Sensing Problems,” PhD dissertation, University of Alberta, 2013.
 P. Kadlec, B. Gabrys, and S. Strandt, “Data-driven Soft Sensors in the process industry,” Computers & Chemical Engineering, vol. 33, no. 4, pp. 795–814, 2009.
 J. Hines, “On-Line Monitoring for Calibration Extension: an Overview and Introduction,” University of Tennessee, Tech. Rep., 2008.