Remaining useful life prediction using motorized system transfer functions

Dear all, I want to predict 'remaining useful life' of a motorized system whos estimated and verified transfer function we calculated 5 years before and now after 5 years we calculated it again. We found significant reduction is system performance w.r.t to bandwidth,tracking response etc.

Now i want to predict when this machine will require major rehauling or will face total failure.

Can anybody guide me....

how i can predict using estimated tranfer functions of 5 years gaps.

Answer 1

[Andy]

One way to do this is to form a ARMA model of the transfer function. If you know the input and output signals, you may should be able to estimate, using a RLS or similar algorithm, the weights of the ARMA model of the transfer function for both a healthy motor and a degraded model. This may give you more of an insight.

http://dspace.lib.cranfield.ac.uk/handle/1826/13665

[arsalan khan]

dear thanks. I have ARMA models and no doubt their weights are different and they showed me degradation but how i can estimate RUL using predictive maintenance toolbox.

[arsalan khan]

Any info how i can get RUL?

Answer 2

[Andy]

Hi Arsalan,

I would start by looking at examples in the Predictive Maintenance Toolbox, there will be an example of something similar to what you want to do.

https://uk.mathworks.com/products/predictive-maintenance.html

Best regards,

Andy

[arsalan khan]

Yes i tried using examples but my RUL estimate (estRUL) always comes ZERO...any specific example where i can find using two transfer functions one (perfect condition) and other one (degraded condition) to get future life prediction.

[arsalan khan]

which specific strategy i need to apply..do i need to I try 'residual similarity model'