What are the limitations of traditional maintenance techniques?
Traditional predictive maintenance is more of a reactive/condition-based monitoring system. Conditions are manually defined by alarms/warnings generated by some standard assumptions. If those alarms are triggered, operators are alerted and actions are taken accordingly. However, it sounds simple but there are significant limitations to such a traditional alarms-based system primarily for two reasons: (1) it is very challenging and human laborious tasks to devise such thresholds, (2) often multi-sensor data is not interpreted simultaneously using advanced ML techniques.
In many instances, the root causes of machine failure are from unknown sources. With traditional PdM, if the root cause is not due to one of the selected sensors, then it’s very difficult to detect it. Another limitation could be the way, the system monitors data. Besides traditional PdM is reactive in nature i.e maintenance is carried out only after the failure episode takes place. Indeed grateful to the platform that is based on advanced AI/ML and big data computing technologies, we can leverage past data to make predictions about future failures.