Set up an impedance divider, the source (50 ohms) into the transducer. Measure the gain, and phase if possible, of this network. (Gain is easy enough, the output should be constant voltage and constant resistance, at all frequencies -- check this by sweeping the output with the generator unloaded, and with a 50 ohm load. It should be flat, and the 50 ohm load should reduce its amplitude by half.)
Phase, you won't be able to measure the "top" of the impedance divider (because that's internal to the generator), but you probably have a trigger output to reference that.
Write down frequency, amplitude and phase in a spreadsheet.
Measure at least a reasonable density of points around the main intended resonance, but also sweep the whole range (say 10kHz to whatever max is, 10MHz say?) to see if there are other modes too.
Also repeat the measurements for different load conditions.
To convert that into a network, a rational approximation method is used to fit the curve. This isn't trivial, but there is an algorithm to do it. Then you can put the network into SPICE.
Tim