Alarm　correlation　analysis　plays　an　important　role　in　telecommunication　network　management.　A　method　for　sequential　pattern　mining　with　time　constraint　was　proposed　to　extract　alarm　correlation　rules.　A　well　defined　mathematical　model　was　presented　and　a　frequent　sequential　pattern　tree　mining　algorithm　(FSPTM)　was　constructed　for　the　sequential　pattern　mining　problem　with　time　constraint.　A　special　data　structure　to　record　the　time　information　pattern　was　also　designed　to　enhance　the　computing　efficiency.　Finally,　the　analysis　result　based　on　four-month　alarm　data　from　one　provincial　mobile　telecom　is　used　to　verify　the　algorithm.