Abstract. A standard classical geomagnetic storm can be divided into three phases, namely initial phase, main phase and recovery phase. We have analyzed the best-fit initial, main and recovery phase duration for 190 large geomagnetic storm events. For the study of the initial phase duration, we have compiled such storm events whose initial phase duration varies in the time intervals of 0-2, 3-4, 5-6, 7-8 and > 8 hours. Similarly, for the main phase, time intervals of 0-6, 7-12, 13-18, 19-24 and > 24 hours have been selected. The recovery phase of storms takes more time, so the time intervals is selected in the varying range of 0-24, 25-48, 49-72, 73-96, 97-120 and > 120 hrs have been chosen. From this analysis, it is found that the maximum number of sudden commencement geomagnetic storms having their initial phase duration lies between 0-2 hours. The main phase duration for maximum number of large geomagnetic storms lies between 7-12 hours. The recovery rate depends on magnitude and main phase gradient of storms and best recovery phase duration lies between 73-96 hrs (3-4) days for maximum number of geomagnetic storms. The occurrence rate of large geomagnetic storms displays a pronounced semiannual variation. Here we have also studied the semi-annual variations of 190 storm events occurring during interval (1986-2007). This analysis indicates that in the first half of annual part (January-July), occurrence rate for large geomagnetic storms is maximum during March and April months. Similarly, in next half (July-December), occurrence rate is higher during October and November. These results show that the months March, April, October and November were more disturbed months and maximum solar activities occur during these months. It is also found that 49.5% of large storm events occur during these months. The semi-annual variation of large storm events shows near approximately a cyclic variation peaking around April and October months. The study of semi-annual variation of storms have an important role in space weather predications.