Within the manufacturing sector, unexpected failures of machinery or parts can lead to significant risks to productivity, profitability, quality and reputation. Historically many companies operated in a reactive mode and waited for equipment breakdown. Over time this transitioned to condition or time-based monitoring. A shortcoming of this approach is that the cost of failures can often far outweigh the cost of routine maintenance or replacement. Most facilities perform change-outs or maintenance in a conservative manner due to uncertainty and fear of failure.

However, an overly conservative approach often results in significant downtime and unnecessary operational expense and often results in excessive spare parts inventories to allow for all possible equipment failures. Recently companies have started to use predictive maintenance systems typically using temperature and vibration data but this is often irregular and lacks sufficient data to be of great analytic value. Now through the use of reconfigurable, real-time, more continuous and diverse data-driven frameworks, we can establish more accurate prediction of potential asset failure and/or dis-improvement maximising asset utilisation, plant throughputs, avoiding potential safety issues, and ensuring product and process quality is maintained.