How to avoid returning failures on a critical vinyl mixer causing long downtime and safety risks?
A vinyl mixer with repetitive failures posed a major challenge for a chemical plant in Belgium. Due to the HSE risks and strict safety procedures interventions are complicated and often resulted in long downtime. This had a high negative impact on the OEE. The FMEA analysis pinpointed the mechanical seals as the root cause of the failures.
When looking at the data with traditional charts and plots, it is not easy to see if any change in patterns occurred before the failure. Using historical data we were able to identify which key factors changed between a healthy period and the period before a failure of the mechanical seal.
The unique P4A machine learning models allowed us to establish a health indicator for the equipment:
In this case we diagnosed that unhealthy conditions could be observed up to 4 months before a failure. This allowed us to configure a simple predictive maintenance tool with alarm thresholds to warn the maintenance team and enable safe operations.
Thanks to the automated alarms, the right preventive action can be taken and overhauls can be scheduled during a planned shutdown thereby increasing uptime, reliability and safety.
Our solution was easily implemented at low cost and thanks to the P4A platform the machine learning models constantly improving themselves.