Real-time Monitoring Solution to Detect Symptoms of System Anomalies

Conventionally, system anomaly detection has been based on setting thresholds and rules. Today, as systems have diversified and the targets of analysis have become increasingly numerous and complicated, the large number of parameters to monitor has made it impossible for analysts to grasp the relationships between them. In addition, resolving these issues with data analysis requires advanced expertise, which hinders data utilization. Fujitsu’s symptom monitoring solution provides a model for detecting symptoms of anomalies. It uses the results of applying machine learning to operational data in normal times to automatically detect any “state different from usual” and then narrows down the target data to identify the cause of the anomaly. This enables high-accuracy and real-time detection of symptoms of anomalies at the site level without requiring advanced analysis know-how. This paper presents a solution that enables real-time monitoring to detect problem symptoms such as information and communications technology (ICT) system failures and suspension of a manufacturing line due to failure of operating equipment.