Event-Based Sampling Architecture and Optimization for Industrial Internet of Things

The proposed paper will present a standardized solution for Industrial IoT, to employ event-based sampling for detecting equipment faults and preventing product impact in large-scale manufacturing. This is important because IoT sensors can generate dense datasets with wide distributions, attributed to varying operating conditions in the fab; sensor measurements summarized without the appropriate context makes it difficult to correlate IoT data with fab processes. To improve our detection rate and minimize false alarms, it is critical to intelligently sample this data using contextual events before applying tighter control. We implemented a standardized solution for event-based sampling which is easily extensible to any Industrial IoT scenario, since it was designed and optimized using industry best practices and evaluated against real fab use-cases. It significantly reduced server CPU consumption and prevented message loss / connection timeouts with other manufacturing systems. Improving our signal-to-noise-ratio (SNR) based on context consequentially helped with tighter process-control, enhanced fab operations, and transformed our detection from reactive to predictive / preventive.