Reform of corporate organization to promote transformation to smart
Digitization of information
More sophisticated decision making
Construction of a system for human resource development , etc.
Construction and sharing of databases
Information visualization and browsing
O&M improvement by predictive detection
Automation of operation and inspection
System Invariant Analysis Technology(SIAT) is part of NEC’s Cutting-edge AI technology that can detect anomaly at an early stage by focusing on the relationships of sensors. SIAT is machine learning technology for time series data, can handle huge amounts of complex data of manufacturing and maintenance sites. Learns relational models from past measured data of normal periods and finds anomalies from real-time monitoring data.
1. Makinga relational model for ”usual state”
SIAT makes a model for ”usual state” from past sensor data of normal periods, you can start analysis with just the data of the usual operating status.
If the time series data is inputed into the SIAT, the relationships between all one on one sensors are created.
Each relationship express the local characteristic of the relevant component. And, by finding a large number of relationships, it is possible to see the characteristics of the entire system from a different perspective than before.
2. Monitoring the relationships
The characteristic of SIAT is not to monitor one sensor value/relationship, but to monitor the relationships between the sensors. You can see the appearance of the relationship between the sensors have been changing, it is possible to detect sign of ”not usual state” at an early stage. Operators can easily understand ”Why and Where the anomaly signs are appeared”. SIAT is explainable AI.
3. Focus on field use
SIAT is easy-to-understand technology in manufacturing and maintenance sites. Various user interfaces are provided assuming that will be used by field operator, user does not need to cover difficult statistics or data science knowledge. Besides, SIAT can be incorporated into the customer’s system or easily linked.
Achievements and examples
Copyright Ministry of Economy, Trade and Industry. All rights reserved.