Extended spherical self-organizing maps for an intelligent management system to automatically detect anomaly conditions

Abstract

This study proposes a safety management system that can be applied to different types of sensors and standards. The proposed system automatically detects dangerous conditions by applying extended spherical self-organizing maps (ESSOM). ESSOM uses algorithms with improved border effects and local minima from existing self-organizing maps (SOM). The ESSOM designed here studied sensor data with different standards and automatically sorted them by condition. The results were compared with those obtained from existing spherical self-organizing maps (SSOM) and from experts. The experiment showed that ESSOM is able to detect dangerous conditions automatically without both real-time monitoring and the mathematical modeling of all possible scenarios because it corrected the flaws in the border effect and local minima in existing SSOM. The results also showed that the system could be applied to safety management systems with various standards and sensor data. ©2013 International Information Institute.