I will present a fast, fully automatic technique for finding multiplets in a microseismic dataset. The technique can be applied in real time to continuous recorded data or to detected event data for a number of three-component receivers and does not require any a-priori information such as P or S wave time picks. We use cross correlation coefficients, evaluated in the frequency domain, as a technique for distinguishing multiplets from non-multiplets. The technique has been applied to synthetic data and an average correlation threshold of 0.9 has been found to successfully identify multiplets from non-multiplets in the absence of noise. With increasing noise the threshold decreases. The technique fails when the noise amplitude is greater than 20% of the maximum signal amplitude. This technique is applied to a three-component passive seismic dataset recorded at an oil field. We have identified a large number of acoustic emission doublets that can be grouped into multiplets, reducing the total number of absolute locations by a factor of 2. 7 larger multiplets reflect the repeated multiple re-rupturing (up to 30 times on a single fault) and significant stress release on approximately 8 different faults. Two major faults dominate the seismic activity causing at least about 1 of the observed events. The amplitude variation within individual multiplets ranges up to a factor of 50, indicating large variations in the ruptured fault segments of individual events.