Abstract: The diagnosis of mine ventilation abnormality （faults） is conducive to winning more time for early warning and repairing the ventilation system, preventing accidents at the early stage, and facilitating intelligent regulation of airflow as well as disaster prevention and relief. The rapid and reliable diagnosis of mine ventilation system abnormalities has always been a challenging issue faced by mine intelligent ventilation, and its premise and foundation lie in the in-depth study of diagnostic indicators and methods. This paper comprehensively elaborates the work carried out and achievements made in the field of mine ventilation abnormality （fault） diagnosis in the past 20 years. It systematically summarizes the research progress of mine ventilation system abnormality diagnosis from three aspects: abnormal diagnosis characteristic indicators, abnormal diagnosis methods, and optimized sensor placement. In terms of abnormal diagnosis indicators, there are mainly air volume characteristics, air volume-air pressure composite characteristics, and air door pressure difference characteristics, showing a development trend from single characteristics to composite characteristics, and then to multi-composite characteristics. In terms of diagnosis methods, supervised machine learning diagnosis methods are mainly used, which rely heavily on the accuracy and quantity of abnormal samples, resulting in low diagnosis efficiency and reliability. In terms of optimized sensor placement, a large number of studies have been conducted on the optimized placement of wind speed and wind pressure sensors from different angles, achieving good results, but the research on the optimized placement of air door pressure difference sensors has just started. Combining the current research status and development trends, it is believed that further research should be conducted on abnormal characteristics based on easily monitored characteristic parameters, integrating other characteristic parameters, and conducting mine ventilation system abnormality diagnosis based on multi-information fusion to improve diagnostic reliability. To reduce costs, maintenance workload, and provide diagnostic reliability, comprehensive optimized placement research on multi-sensors should be carried out. On this basis, aiming at the difficulty of obtaining a large amount of abnormal （fault） data, supervised machine learning diagnosis methods require continuous retraining and learning, and online diagnosis is difficult. Therefore, research on diagnostic methods such as graph theory, rule-based expert systems, and unsupervised learning should be conducted to improve diagnostic efficiency and reliability.