Experimentally, I will use accelerometers to measure the acceleration during normal operation and of course you can measure that in different axis directions. Also, as Daminain said above, I will model whatever you want to measure vibration to such as model the blade or the turbine, etc., and then have the proper material properties and calculate the natural frequencies of the system. This is will give you the natural frequency modes and values. Next, you can apply whatever speed, or any boundary loading conditions on your model such as loads, speeds, etc to find the forced frequencies. This is can be compared to experiments to correlate. If the design is good, natural frequencies shouldn't match the forced frequency of the loadings.Also, natural frequency should not be so low, otherwise this an indication of bad design.
What do you mean by load?? If you want to identify the frequency of excitation you have to know the construction of the object to determine excitation deriving from motor, gear, shaft, blades (BPS - blade passing frequency)..... Thank's to those information the spectrum interpretation will be easier, if you made the measurements during operation. You can also perform experimental/numerical modal analysis to identify natural frequencies. If you give more details, maybe I will be able to give you more specified answer.
Vibrating frequency and amplitude, for all above items can be measured by appropriate, variable inductance or capacitance type transducer, where inductance/ capacitance is varied by vibrating core/distance in the transducer(LVDT), supplied by constant voltage source, gives variable current or secondary voltage , and can be measured with C.R.O.
Experimentally, I will use accelerometers to measure the acceleration during normal operation and of course you can measure that in different axis directions. Also, as Daminain said above, I will model whatever you want to measure vibration to such as model the blade or the turbine, etc., and then have the proper material properties and calculate the natural frequencies of the system. This is will give you the natural frequency modes and values. Next, you can apply whatever speed, or any boundary loading conditions on your model such as loads, speeds, etc to find the forced frequencies. This is can be compared to experiments to correlate. If the design is good, natural frequencies shouldn't match the forced frequency of the loadings.Also, natural frequency should not be so low, otherwise this an indication of bad design.