Interesting question (that was also discussed in PRé's LCA list for example) - a normal LCA model is surprisingly stupid and does not consider time; a life span of products is interesting in two ways, for an LCA:
a) possibly, resource demands and emissions are distributed over time, and an emission e.g. in 1000 years probably has different specific impacts than an emission today. This is well recognised in costing, with a discount rate that recognises different "value" of costs in different times. This, however, is also not really recognised in common LCIA methods and thus in practice often ignored. For some processes, like landfills, common models sometime have a simple, constant emission rate over time, and if you consider that waste is in the landfill until eternity, you would have infinite emissions, which is of course not realistic, and thus it is common to cut the emissions after some time, e.g. 35 years.
b) for a longer living good, such as a pump, you need to "allocate" a part of the life span to the functional unit; if the pump pumps overall 1E6 m³ over its life, you need 1E-6 pumps per m³ pumped fluid. You see that you do not really need time / life span itself, but rather the product (pumped water amount) that is delivered over the life time.
You can of course combine different long-living goods in one LCA, following this principle.
I hope this helps!