Applying machine learning (ML) on operational data stored in process historians (e.g. OSISoft PI system) has considerable potential to reveal critical information that supports decision-making at different levels of an organization. In particular, ML can significantly reduce operational costs in the oil and gas industry, and at the same time, improve robustness and safeness of industrial processes.
Implementing ML-based applications using operational data is still costly and challenging due to the special nature of the data. Technical decisions in the data pipeline might affect the suitability and quality of operational data for ML. This might include insufficient information about the data (metadata) and the assets associated with each data stream.
- Create guidelines and training material to make operational data suitable for ML-based digital services
- Implement an ontology-based digital service to help data scientists understanding and preparing operational data for ML-based digital services
- Improve existing services to facilitate the exploitation of operational data by ML-based digital services.
- Reduced development costs and improved precision and quality of ML-based digital services applications
- Facilitation of the interpretation, location, access, integration and sharing of operational data stored in process historian.
Versatile ML-based digital services will help oil and gas operators reduce operational costs and improve robustness and safeness of their processes through:
- Highly precise prediction of failures of critical systems
- Optimization of maintenance operations
- Faster and more accurate decision making
- Decrease production costs
- Reduction of effort and time to create ML-based digital services.