It’s an unfortunate fact of reality that no-matter how advanced the equipment, there will come a time when the machine unexpectedly fails and your factory experiences losses caused by unplanned machine downtime. The costs of this can vary dramatically; from over $1.5M per hour in automotive manufacturing to thousands per hour in the Fast Moving Consumer Goods (FMCG) industry; no-matter the cost it’s still extremely undesirable.
Even machines cared for under the best maintenance regime will experience unplanned failure at some point in their lifetime.
Preventative maintenance is often expensive and generally a flawed concept. Moving to predictive maintenance is ideal from a cost/reward point of a view however it must be supported with effective condition monitoring at a minimum. Condition monitoring, either offline or online fundamentally has scalability issues as it requires expert manual interpretation, following a data collection period (with costs significantly increased if manual data collection is required). There’s a limit to how much condition monitoring data an engineer (or even a whole team) is able to effectively interpret, with dramatically diminishing returns as the number of assets under analysis increases.
Condition monitoring also only tells half the story as it will only reveal the current state of the machine; for the future you need prognostics.
Prognostics is the science of predicting when a machine will stop being able to perform useful work – this could be because of an unacceptable variance in tolerance or an outright mechanical failure. The important thing is that it allows maintainers to improve OEE and reduce unplanned downtime by giving them the knowledge of when a machine will fail. This is calculated as Remaining Useful Life (RUL).
With an accurate prediction of RUL, maintainers are better able to prioritise which machines need maintenance and in what timeframe; significantly reducing the risk of machine failure and avoiding costly over-maintenance. Think of it as another tool to help boost the effectiveness of maintainers.
Prognostic models giving RUL have traditionally been difficult to calculate and have needed talented data scientists to implement effectively on a bespoke basis. The high costs of this approach have harmed widespread adoption. Thankfully with the advent of advanced machine-learning and inexpensive cloud-computing, prognostics is becoming more accessible and in-turn predictive maintenance is achieving scalability.
The key enabler of predictive maintenance
Senseye is the market-leading cloud-based product to automatically analyse condition monitoring data and calculate prognostic models. As an introduction to this detailed topic, we’ve put together a free white paper.