Think of a complex system with very expensive
parts. We can't risk running into failure as it will be
extremely costly to repair highly damaged parts. But
more importantly, it's a safety issue. This is why
numerous organizations attempt to avoid failure
beforehand by performing regular inspections on their
equipment. One big challenge is to determine when to do
maintenance. Since we don't know when failure will
occur, we have to be conservative in our planning. LTSM
can be used to predict the remaining useful life. But if we
schedule maintenance very early, we will end up wasting
machine life that is still usable, and this will add up to
our costs. However, if we can predict when machine
failure will occur, we can schedule maintenance right
before it. Recurrent Neural Networks can predict when
this machine failure is bound to happen. Predictive
maintenance lets us estimate time to failure. It also
pinpoints problems in complex machinery and helps us
identify what parts need to be fixed. This way, we can
minimize downtime and maximize equipment lifetime.
PREDICTIVE MAINTENANCE USING LONG SHORT-TERM MEMORY ALGORITHM ON SENSOR DATA
