Compressor management for an upstream major, with an engineering partner
Real-time compressor condition monitoring, anomaly detection and root-cause recommendations proposed with an engineering partner, on an approach proven on four offshore platforms for two operators.
Challenge
An upstream major wanted to get from compressor sensor data to meaningful action: alarm management, connecting the right sensor to the right operating issue, and root cause of abnormal conditions, with too few failures to train failure prediction directly.
Solution
With Worley, Arundo proposed a three-phase delivery: use-case verification with the operator's experts, configuration and modelling in two-week sprints using distance-based clustering or auto-encoder models trained on normal operation and mapped to operating modes, then testing, documentation and training.
Impact
An operational compressor-monitoring solution with model output by API, email notifications and a configured user workflow; the two-weeks'-warning compressor case served as the reference.