1,984 multivariate time-series instances capturing 8 types of undesirable events in offshore oil wells. The data consists of sensor readings for pressure and temperature variables, such as P-PDG and T-TPT, sampled at 1Hz across real, simulated, and hand-drawn scenarios.
Use Cases
- Train anomaly detection models to identify 'Severe Slugging' using the P-TPT and T-TPT sensor columns
- Develop classification algorithms to distinguish between 'Real' and 'Simulated' event instances
- Analyze pressure drops in the P-PDG variable to detect 'Spurious Closure of DHSV'
Strengths
- 1,984 labeled instances across 8 event classes including 'Severe Slugging' and 'Gas Kick'
- Features 5 primary sensor variables: P-PDG, P-TPT, T-TPT, P-MON-CKP, and T-JSC
- Categorizes data into three distinct sources: real occurrences, simulated events, and hand-drawn patterns