Apparatus and Method for Anomaly Detection Using Multivariate Heterogeneous Time-Series Data

摘要

The present invention is implemented with a data acquisition unit that acquires a plurality of time series data generated in a process process, and divides it into process data acquired in units of process cycles and sensor data acquired regardless of the progress of the process, and a pre-learned artificial neural network A first auto-encoder that encodes and decodes process data to obtain restoration process data, a second auto-encoder that is implemented as a pre-trained artificial neural network to encode and decode sensor data to obtain restoration sensor data, process data and restoration process A restoration error acquisition unit that calculates the process restoration error and the sensor restoration error from the difference between the data and the sensor data and the restored sensor data, and a time period in which both the process restoration error and the sensor restoration error indicate anomalies as the final abnormality occurrence period It is possible to provide an abnormality detection apparatus and method capable of accurately detecting an abnormality occurring during a process, including an abnormality detection unit.