A global oil and gas company obtains actionable insights with a machine learning predictive analytics solution.
Our client, a Global 10 oil and gas supermajor has a top-level initiative to reduce ongoing operational costs through advanced analytics and predictive machine learning. Significant challenges include developing machine learning models for multi-dimensional sensor data for subsea electrical systems and oil and gas pipelines. Additionally, our client faced challenges in leveraging a third-party machine learning solution from SparkCognition which included long turnaround time and depth of analysis.
The client engaged Apex for exploratory data science and the development of neural network models for predictive forecasting. Machine learning and statistical models were developed both to predict failures and to provide more detailed identification of key sensor data that could be utilized for additional forecasting.
Improved detection time for system failures and accelerated model development
The engagement delivered the following functionality:
- Ongoing development of a deep learning neural network for unsupervised anomaly detection and supervised predictive forecasting of equipment failure
- Ongoing consultation and exploratory data science for evaluation of machine learning model results and model feature importance
Our solution provided actionable business insights and operational machine learning model data in support of the client’s engagement with SparkCognition. TensorFlow state-of-the-art internally developed neural network models were developed by leveraging the client’s internal high-performance computing systems. As an ongoing effort, this project is enabling faster turnaround time on operationalizing data science insights and growing to support additional large-scale upstream oil and gas systems.