Success Story
How an oil & gas company revolutionized anomaly labeling and achieved 10x faster data preparation for supervised learning
Labeling multi-year well data and identifying anomalies was time-consuming & labor-intensive
AT A GLANCE
Challenge: Labeling sensor data from multiple wells spanning multiple years
A leading oil & gas company was facing the challenge of efficiently labeling sensor data from numerous wells over several years. The task involved categorizing various operational stages and distinguishing anomalies, presenting a time- consuming and labor-intensive process. The existing approach, reliant on manual investigation of each event without interactive labeling capabilities, hindered efficiency and accuracy.
Solution: Self-service tool for manual and automatic labeling for SMEs
Integrated seamlessly into the company’s infrastructure, Visplore empowered engineers to ingest and analyze extensive operations data spanning multiple years. Leveraging advanced pattern-search and rule-set algorithms, Visplore automated the segmentation and labeling of operational stages such as start-ups, transients, stable periods, and shutdowns, liberating engineers from manual tasks and enabling swift decision-making. This allowed engineers to compare all such events and decide which instances to use as training data for their model.
Furthermore, Visplore facilitated interactive labeling of well statuses by subject matter experts, revolutionizing the process. This approach not only accelerated labeling tenfold but also empowered engineers to delve deeper into their data, gaining invaluable insights into anomaly events and operational behavior (see image*).
Results: Time savings and clearer insights for model improvement
Engineers reported not only a remarkable 10x increase in labeling efficiency, freeing up valuable time for strategic analysis and innovation but also an increase in insights gained from the data. For example, automatic segmentation revealed extra stages that were previously unknown to the engineers, enriching the training dataset and enhancing model accuracy. Additionally, Visplore’s visualization capabilities revealed nuanced insights into anomaly events, driving continuous improvement and operational excellence.
Highlights
