Predictive maintenance: production plant under control
With predictive maintenance, failures can now be predicted before they occur thereby improving production plant performance.
Sensors and data are at the heart of predictive maintenance. “In terms of its approach, this is a completely different system to preventive maintenance, which involved changing machine parts based on their supposed state of wear”, explains Bernard Cusenza, manager of Actemium Maintenance Bourogne. In predictive maintenance, the analysis is based on data supplied continuously by sensors indicating a part’s actual state of wear. As different parts are subject to varying loads and stresses, the wear will not be the same. The sensors provide precise information about each part’s wear. Concretely, this means changing a part after 2500 hours’ use for example, whereas under preventive maintenance, it would have been changed after 2000 hours whatever the stress or load.
With the technology available on the market today, we can monitor everything and imagine almost any scenario.
“Ultimately, this means we are reducing maintenance costs and we are able to plan operations to minimise production outages”, says Patrice Fortea, manager of Actemium Maintenance Belfort. “With these sensors, we are able to fully manage flows and production based on the status of the plant being monitored.”
Using sensors intelligently
Sensors are one of the key components of predictive maintenance, but they must be used appropriately. “With the technology available on the market today, we can monitor everything and imagine almost any scenario. However, we must not lose sight of the fact that we really only need the most relevant data to predict and manage failures. Of course, we can install sensors everywhere, but then we risk being inundated with data”, warns Bernard Cusenza. “That’s why we conduct an analysis of the kinematic chain for vibration analysis so that we can determine the strategic installation positions for the sensors.”
Counter-productive, the massive use of sensors also comes at a cost, even if it has fallen considerably. “As we can’t afford to put sensors everywhere, the choice of which machines will have sensors installed is very important. Clearly, the focus must be on three types: plant that is strategic for our clients, that requiring expensive spare parts, and older equipment for which spares are sometimes no longer manufactured”, explains Patrice Fortea.
The three approaches of predictive maintenance
Predictive maintenance mainly involves three measurement technics :
- Acoustic: ultrasound is used to measure the state of the equipment
- Thermographic: sensors measure temperature to detect abnormal levels of heat
- Vibration analysis: the machine’s vibrations are analysed to identuify any operation faults
Data analysis, a key aspect of predictive maintenance
It is the data transmitted by the sensors installed on the production plant that is used to analyse any discrepancies that predict a failure.
By comparing the analysis of these parameters against the plant’s historical data, it is now possible to programme production shutdowns rather than have to deal with failures when they occur. “Having a historical record of a machine’s failures, experience feedback or having performed an FMECA will, for example, help us specify the points to be monitored”, explains Bernard Cusenza. “We define alert thresholds based on our interpretation and our knowledge of the plant”, adds Patrice Fortea. Once the threshold has been reached, the work can be scheduled. The relevant part is replaced just-in-time. This principle is found on a large scale in smart industry. As Patrice Fortea explains, “We can draw a parallel between predictive maintenance and the steps taken by our clients to improve just-in-time processes introduced in their production organisation. Today, the implementation of a maintenance strategy is also based on just-in-time principles insofar as failure detection and maintenance work are concerned.”
Limited spare parts inventory, fewer unplanned production outages, and improved production plant performance and lifespan: predictive maintenance makes it possible to work towards the goal of zero outage and zero failure