I work in a manufacturing company where we still rely primarily on traditional maintenance practices, meaning that equipment is usually serviced only after a failure occurs. However, unplanned production downtime is becoming increasingly expensive, and we are more frequently considering the introduction of systems for monitoring machine conditions and predicting potential failures before they happen. This made me wonder whether any of you have already implemented predictive maintenance using sensors, data analytics, or the Industrial Internet of Things

How much were production downtime and maintenance costs reduced after introducing such a system, and did the investment prove to be worthwhile for small and medium-sized manufacturing companies?

Any advice or practical experience would be greatly appreciated.

A basic maintenance schedule will pay for itself pretty quickly. You don't need fancy predictive models unless you are really fine tuning the program. The military uses a basic time based program, e.g. grease all parts of a tank suspension every X kilometers of use. Back in the '60s I worked at the steel mills in East Chicago, IN and it was part of my job to do certain lubrication activities once/shift, e.g. attach a grease gun to the grease zerks on a compressor and pump until clean grease came out at the end of the bearing.

I once worked at a mine in Colorado which ignored this sort of thing and had a philosophy of "run it till it breaks and then fix it the fastest and cheapest way possible and if it can't be fixed replace it with the cheapest used unit you can find." Needless to say, their production was poor and their costs high. If they had hired a maintenance guy to go around to the various mine locations and grease the equipment and tighten things up they would have paid for his wages many times over.