Once you have a reasonable collector current, which looks as if it should be a bit over 1mA, for the values you have, the next step is to give the circuit a nudge to try to start it. In the real world noise voltages should do this, but a simulation may need an actual kick.
I suppose that you are doing a Transient analysis using something like PSpice. You could try putting a pure current source in parallel with the crystal, and applying a small pulse, say 1mA high for a pulse length of half the crystal period (33ns for a 15MHz crystal).
Oops - nanoseconds, not milliseconds!