Turbine Failure? It's Not An Option.
Overview
The customer, based out of UK, is a Fortune 100 energy solutions company committed to addressing the energy challenges of today and tomorrow. The company manufactures gas turbines and deploys at various customer locations and has SLAs for uptime of the turbines.
The Challenge
The client manufactures a comprehensive lineup of gas turbines for a wide range of applications from small turbines for industrial use to large ones for nuclear power generation. Irrespective of strict in house testing and high quality management, processes of these steam turbines, the Client has been incurring huge losses from machinery breakdown. As a result, there were huge costs incurred (that include penalties for down-time), longer wait times involved in repair, resulting in a wasted time, opportunity and efficiencies due to these business interruptions.
Goals
To mitigate such shortcomings, the Client wanted to develop a scalable real time solution for early detection of cracks in gas turbines. To develop a highly accurate and dependable solution that can help in predicting possible cracks/fissures at least 6 months in advance to minimize, if any, downtimes related to turbine failure.
The Strategy
- Come up with a solution with the use of data-driven methods, using AI/ML
- Perform a Feature Engineering exercise which is vital to gauge the connection between the data attributes
- Create a program to extract the identified features from the data automatically
- Fine-tuning the programme to estimate the possibility of a crack in the blade
Result
- Early detection of cracks in gas turbines was an extremely complex engineering use case that Techno-Comp solved with high accuracy, saving millions of dollars every year.
