We do this periodically for JVMs that are prone to going "sick" but it is a bit fiddly
Create a calculator against a java metric that should always report - e.g heap bytes in use
Create a metric grouping from that calculator
Create an alert from that metric grouping
Alert if value < 1, as Heap should always have a positive value, so if is no longer reporting the calculator value will be 0
Typically we find that these sort of jvm's will be javacoring at the same time anyway, so we have a default core file detector running via EP agent as well.