Abstract:
Modern technical systems such as power plants are equipped with diagnostic systems. But these systems are imperfect, and accidents still happen. Accidents can lead not only to economic losses but also to socially frightening consequences, such as human-made catastrophes. Another one, that seems to be a typical, problem for Nuclear Power Plant is the excessive duplication of safety systems, which increases the cost of the NPP itself. The solution could be an advanced diagnostic system. Diagnostics as science can be divided into the main three parts: the first one is the monitoring of technical conditions; the second one is finding the root cause of anomalies; the third one is the forecasting of the future state of a technical system. The developed framework can be used to solve all these diagnostics tasks.
Keywords:technical systems diagnostics, machine learning, deep learning, time series analysis, anomaly detection, time series forecasting, data preprocessing, framework.