BUILDING UP AN INDUSTRIAL DATA PLATFORM FOR ASSET MAINTENANCE
Many industrial companies have faced the challenge of implementing predictive maintenance programs. Mature machine learning tools are already available, and inflated marketing expectations have increased top management sponsorship. But the crude reality is that predictive maintenance projects have still shown very little impact in corporate manufacturing environments.
The big barrier to achieve the expected results is the data integration problem. Different equipment and assets require different approaches to predictive maintenance that need to be properly orchestrated. Beyond the asset diversity problem, predictive maintenance solutions need to be integrated with the overall legacy maintenance process and systems.
HOW WE HELPED
Galeo has worked with several customers to successfully design and consolidate Predictive Maintenance programs on the appropriate data platform infrastructure. Our work with them has been based on a modular approach based on information reuse and integration. This included:
- Data capture infrastructure – including sensors and gateways – with the proper abstraction layers for management and data flows.
- Data governance and exploration, with automatic catalog capabilities. This component enabled data exploration and discovery for eventual reuse, making operational data available to predictive systems in a fully-contextualized way.
- Establish integration mechanisms, to quickly prototype complex asset maintenance systems by combining different specialized puzzle pieces: A predictive system for rotating equipment may be combined with a corrosion prediction system chemical process plant, for example.
The modular approach we followed, whose core functionality was oriented to information integration and modular combination of predictive solutions, has confirmed the possibility of accelerating the delivery of a predictive maintenance program. System development gains traction as puzzle pieces shorten time to value, improve the agility to pivot failed approaches, and eliminate the duplication of low-value efforts related to data infrastructure engineering.