To address the challenges associated with acquiring fault samples from mine ventilation systems and the scarcity of research on online diagnostics, we propose the OCISVM model, which integrates a one-class support vector machine (OCSVM) with incremental learning (IL) to facilitate real-time fault diagnosis based on sequential sensor monitoring data. During the offline phase, only normal sample data collected by sensors is utilized to construct the classification hyperplane. In the online detection phase, leveraging incremental learning principles, we update the classification hyperplane by incorporating a Delta function while assimilating new samples. Ultimately, we establish a test threshold for the model based on predefined criteria and conduct online fault diagnosis for test samples. This method has been applied successfully to a mine ventilation system. Experimental results indicate that our approach is effective for diagnosing branch faults in mine ventilation systems, achieving an accuracy rate of 97.5%, with diagnostic times at millisecond levels and enabling real-time diagnostics. Compared to conventional classification methods, this approach demonstrates enhanced stability when handling imbalanced datasets. Furthermore, relative to the recently introduced CMA-ES model, improvements in AUC, F1 score, G-mean, and ACC metrics are observed at 2.1%, 3.3%, 4.7%, and 1.9% respectively; additionally, testing time is reduced by an impressive 99.3%.