A Distributed Modeling and Monitoring for High-dimensional Non-stationary Data in Industrial IoT Scenario

Industrial Internet of Things (IIoT) devices often face interference from complex external environments, resulting in high-dimensional operational data exhibiting non-stationarity. This paper introduces a distributed monitoring framework designed for such high-dimensional non-stationary processes within IIoT settings. Central to the approach is a novel decomposition method utilizing "degree of cointegration" to partition the high-dimensional variables into multiple distinct blocks. Subsequently, each sub-block undergoes independent monitoring employing cointegration theory-based techniques. The method’s effectiveness is validated through two real-world IIoT case studies. Results demonstrate that the proposed framework surpasses current benchmark methods and significantly aids in fault isolation.