Maintenance & COMPREDICT

Corrective Maintenance is understood as the action taken after a failure: Maintenance is carried out after a part has failed.

Important: Parts must be safe to fail (failure is easy do observe, does not harm other components and user is not at risk)

• As long as possible service life -> Lowest service cost

• Very Simple

• Low reliability —> Considerable and unplannable system downtime

• Safe to fail design must be adhered to

In the case of preventive maintenance, parts are replaced before a failure is expected, no matter the current condition.

• Near zero downtimes

• —> Reduced risk during operation

• —> High efficiency of the equipment while in use

• Unefficient Maintenance —> Relativly high cost

• Relativly high Maintenance time —> Less usage of the equipment

• Detailed planing and modelling needed (otherwise unsafe failure or unreasonably high costs)

The condition of to be maintained components is monitored, and replacements are carried out accordingly.

• More efficient use of resources then with Preventive Maintenance

• More efficient use of equipment then with Preventive Maintenance

• Condition Montioring is needed.

• Unplanned downtimes more likely

The condition of the system is indirectly assessed via the operating parameters and a system model —> Failure is predicted and maintenance carried out accordingly.

• Most efficient use of the equipment

• Most efficient use of resources

• No direct condition monitoring necessary —> Standard parts and retrofitting possible

• Downtimes can be planned, because State of Health is constantly assesed and failure during planned downtimes is predicted

• Good system model necessary

• Unplanned downtimes still possible (However this is influenced by the quality of the model)

The Motivation for Predicitve Maintenance derrives from the combination of the ideal use of the given ressources and an efficient use of the system:

If the system behaviour is understood well, an analytical white box model can predict failures. However in real world applications the systme behaviour is not known completly, therefore most often grey box models are used.

If the system behaviour is not udnerstood, self-evolving, data-driven black box models can be used.

CM gives the current State of Health (SoH). Mostly after a certain point of oberserveability.

DP on the other hand not only gives the current SoH, but also gives a future estimate how the SoH is going to evolve, and can therefore give a Remaining Useful Life (RUL) estimate.

Mean Time to Failure

Mean Time Between Failure

Mean Time to Diagnosis

Mean Time to Repair

SM: Safety distance

sigma_W: Load capacity

sigma_B: Mean load distribution

s_W: stochastic spread of the load capacity

s_B: stochastic spread load distribution

With

sigma_W: Load capacity

sigma_B: Mean load distribution