Big data technology can be used to identify safety-critical events in many systems that generate time series data. This is important because detecting hazard behaviors and determining the duration of failures based on analysis of the time series data ...
Big data technology can be used to identify safety-critical events in many systems that generate time series data. This is important because detecting hazard behaviors and determining the duration of failures based on analysis of the time series data promote system reliability. A Cyber Physical System (CPS) requires a very high level of reliability because it can lead to economic losses, as well natural or life crises if a system error or failure occurs. Therefore, the CPS has to consider time series features and related big data technology for its functional verification in order to detect errors and failures.
In order to make the CPS more reliable, this research included three main aspects. First, the study proposed a multilayered CPS network architecture and a safety-critical CPS node. The proposed CPS network architecture includes components that are in a single CPS layer with the big data processing infrastructure layer, and based on an MDA approach, this research designed meta-models and models for the components that require the safety-critical feature. This research also proposed a transient state, discrete state, and safety state for the CPS state machine, and adopted a reporter, and stabilizer for the safety-critical operations of the CPS node.
Secondly, this research proposed a big data analysis architecture that can interact with a CPS layer using a Hadoop framework. The front-end layer of the architecture collects CPS data, and also provides useful information or emergent warnings to the CPS layer. The back-end layer analyzes the CPS big data and locates hidden information. This research adopted two strategies in the architectural design: purpose-oriented interactive processing and domain actuatable processing.
Lastly, this research proposed an appropriate parallel mechanism that can detect safety-critical events in temporal patterns. The mechanism works on the Hadoop framework and detects scattered partial events. Then, it sorts the partial events to ensure their successiveness in the secondary sorting process and generates time series events according to the patterns. To promote accuracy and reliability, this research used DTG big data as an actual time series dataset and preprocessed it in phases that corrected, optimized, and concatenated the data. Then, this research used experiments to verify the feasibility, accuracy, and high performance of the proposed mechanism.