Context and challenges
This client is a leading group in the nuclear industry. At one of its nuclear power plants, hydraulic water rakes were installed to clean the strainers on seawater collection pump stations.
These rakes feed the cooling water circulation pumps (secondary cooling systems) and the ESWS pumps (redundant raw water - primary cooling system). These latest-generation rakes are cleaner and were brought into service recently, but the client has no feedback in terms of their operation.
The client therefore wished to implement monitoring of these rakes, to track movements and detect anomalies, in order to balance maintenance operations while ensuring the operational effectiveness and good ageing of the facilities. Such anomalies may be: container failing to tip, travel not completed, vibration during movements, blockage during movement, engine overheating, rake stopped, no power supply to engine, etc.
Therefore the client entrusted Assystem to design and implement a data science solution to deliver predictive maintenance, including Artificial Intelligence to track and monitor hydraulic rakes activity and plan maintenance operations ahead of time.
- Design and installation of an independently-powered acquisition chain (of sensors) which records the tri-axial position of the rakes every second (speed of movement of container approx. 5 cm per sec-1).
- Automatic data retrieval and transmission to client server.
- Processing and formatting of data, including information on weather, tides and all other maintenance data, on the Saagie Big Data platform:
- Implementation of a procedure to detect anomalies and diagnose failures.
- Creation and update of an HMI to monitor the operation and state of health of the system.
- Creation of a recommendation module using Artificial Intelligence, taking into account the results of the prediction model. AI is used to predict failures, taking into account several data sources.
- A daily summary report with recommendations is sent to the client by e-mail.
- Optimised maintenance actions and predictability of operations.
- Rapid retrieval of information on the activity and maintenance requirements of the rakes with real-time transmissions.
- The solution that can be transferred to other strainer rakes in nuclear power plants next to the sea or rivers, or other equipment (pumps, pipes, engines).
- Easy installation of the sensors and interfaces with rakes due to the non-intrusive character of the equipment, which requires no wiring.
- Optimised time spent on maintenance operations through the elimination of equipment requalification phases.