Key Technologies on High-Quality Event Detection through Big Data Mining in Industrial Cyber-Physical Systems

Integrating cyber aspects of wireless sensor networks and physical aspects of structures (e.g.,
industrial plants, buildings) in an industrial cyber-physical system (ICPS) can be promising.
However, such an ICPS system produces high resolution and variety of data, which is big data. As
a result, guaranteeing high “quality of event detection (QoE) is quite challenging due to low
“quality of data (QoD)” collection, big data transmission over large structures, and resource
constraints. In this proposal we consider the problem of detection of structural health event through
data mining in the ICPS. First, we propose to study the “differential sensor pattern” mining in a
parallel and distributed manner. This will extract a pattern of sensors that may have event
information. Second, we will study QoD-aware data collection. Through this a sensor mines data
based on signal importance at the time of acquisition and before transmission. Third, we will study
mining “interesting frequency content pattern”, which treats high frequency and low frequency
content from sensor acquired data differently for high QoE so that an uninteresting frequency
content pattern is dropped. Finally, we will study the data mining based on “compression”, which
mines the event-sensitive data through an “application-specific data compression” technique. This
makes decision on the “event-sensitive data” for reliable wireless transmission in the ICPS. The
research outcome will create reasonable choices for civil/structural/industrial engineers and
respective communities to ensure economic benefit and public safety in functioning structures.