Used to do it regularly for certain pieces of equipment ( expensive stuff or stuff that ran in a bad environment - street sweepers, fire trucks, etc. ) in the fleet I was maintaining. It was good for setting maintainence intervals and spotting potential problem areas early - but, it was pretty useless without having a history to look at ( trend analysis ). Of course, you could always spot a catastrophec failure, but you usually didn't need an oil analysis for that
. To get full benefit from oil analysis, you need to start by supplying a virgin oil sample as a base line ( TBN ). Then keep on a very ridged sample interval and don't change anything. A lot can screw up the readings ( oil usage between intervals, any engine repairs, sealing materials used, gasket materials used, intake filter leaks / clogs, etc., oil additives, even adding oil between samples of a different brand ) and if you are the least bit paranoid, your gonna be doing a lot of unnecessary tear down work due to miss-reading the sample. A high silicon reading can mean several different things - one as simple as a leaking air filter or blowing dust. I won't even get in to fuel dilution.
Also keep in mind that when there is a part failure in our type of invironment, there is seldom an early marker, at least not between "normal" maintenance intervals. In short, oil analysis has it's place, but it aint here. JMHO.