The proper way I think is to include if in simplified way the equipment and its structural link to the building
If you need to forfeit the entire building design, UBC 94 section 1632 was dedicated to nonbuilding structures but makes an interesting exception for rigid equipment of natural period less than 0.06 secs, for which a V=0.5·Z·I·W is directly given.
I propose you may also use the following procedure, that engages the evaluation of two simple impact factors, one for the soil-building pair, and another for the building-equipment pair.
Q=40 for reinforced concrete structure
Q=45 for presstressed concrete structure
Q=60 for bolted steel structure
Q=100 for welded structure
r1=fN_earthquake / fN_building
The natural frequency of the earthquake can be read from the strong motion record; there uses to be 3 or 4 strong motion responses up to 5 Hz, so maybe 3 Hz can be typical when unknown.
The natural frequency of the building can be read from lookup tables for the different types of buildings; the spanish seismic code NCSR-02 renamed NCSE-02 has a set of formulas to establish a useful guess of such natural period (inverse of natural frequency).
Impact_factor1=1/(sqrt(1+r1^4-r1^2·(2-(1/Q^2))))
not to be taken less than 1
Now the exact same process is repeated for the impact factor between building and the equipment atop superstructure.
r2=fN_building / fN_equipment_superstructure
The natural frequency of the building can be read as above. You may need to calculate or guess the natural frequency of the equipment, but you gain that you need not to model an entire building for a roof job just out of this thing.
Q normally now that of bolted or welded.
Impact_factor2=1/(sqrt(1+r2^4-r2^2·(2-(1/Q^2))))
not to be taken less than 1
Now if bga is the basic ground acceleration at pavement level, considering the following horizontal force at the center of gravity of the equipment must be a reasonable guess of a seismic force there:
F=W·bga·Impact_factor1·Impact_factor2
where W is the tributary weight of the equipment at its c.o.g. (you may subdivide the forces if complex in shape, or better subdivide equipment by support structures).
This is based on Bolton's statement of the calculation of an impact factor:
Structural Dynamics in Practice
A guide for Professional Engineers
Arthur Bolton
McGraw Hill 1994,
not an earthquake seismic design text.
On the other subject, still to add that...
precisely what the codes are doing when giving some consideration to the subsoil is to mimick through their method the amplification of accelerations that happens through the soft soil layers, so for normal buildings I have scarce doubt that the proper surface value is what needs to be used. To proceed otherwise I would think we would need to have a building rigidly tied to the bedrock, and likely quite deep, for when superficial no code has cared to express correction to the proposed value (other than to take the proper type of soil).
But if you were in such situation of deep bedrock and building rigidly founded there the soil itself would become in its vibration load for the building, and you would have first to have the acceleration at the bedrock derived from that at the surface (which can be made just by an impact formula similar to that used above). Then it would appear the problem of how to realistically (even if in simplified ways) model the earthquake hammering on the basement walls, for it can heavily affect the structural behaviour. An inmediate approximation is to model some solid of soil around the basement hole in elastic terms, that if scandalously incorrect, we use to practice everytime we calculate halfspace pressures etc. A nice problem waiting for people to add knowledge.
