I always have a geotechnical engineer or engineering geologist provide suites of time history records that they consider to appropriately reflect the seismologic conditions at the project site.  The building code should provide guidance on how many sets of time history records are needed.

Hy,

it is a spectrum I want to use to simulate the behaviour of a turbine (especially the bearings) inside a building and the time history has to build according ASME.

Michael

As Taro said, you need an appropriate ground accelerogram suite, reflecting the seismologic conditions at the site, and the use of the building code rules to select and “scale” this accelerograms to the code requirements.
In the meanwhile, you can use existing recorded accelerograms to “build” their associated response spectrums and check which one reproduces more accurately the spectrum you have. You can use a single accelerogram or a “linear combination” of several recorded accelerograms, properly scaled to obtain the closer look to your spectrum.
To build the spectrum, you can use your current finite element software, modeling (in a single model) a series of linear vibrators of different frequencies in the range of interest (simple spring-mass 1DOF models, all of them in the same model) submitting to time-hostory analysis and registering the maximum acceleration of each one. These values will be the ordenates of the spectrum at the frequency of each vibrator.
Don´t forget to combine accelerograms of several "main frequencies", that is with predominant response at different frequencies.
You can obtain several recorded accelerograms at http://training.fema.gov/EMIWeb/nonlin.asp.
I hope it may help you, at least in the beggining.
Regards.