Multiple Profiles Sensor-based Engineering Data Processing

Abstract

Nowadays big data is widely adopted in industry field. In an advanced manufacturing system hundreds of sensors are deployed to collect key variables for system performance and the real-time data would be used for further monitoring and anomaly detection. However, there are many challenges for applying the sensor-based data directly, including the profile data has unsynchronized different length for different samples, the existence of obvious long-term drift, strong correlation of sensor clusters and the particular feature extraction. To solve these problems this invention presents a multiple profiles sensor-based engineering data processing system, including (1) preprocessing the signals to align the data and remove long-term drift, (2) clustering the sensors which have strong correlations, and (3) extracting particular features from different sensor clusters.