Abstract: Based on gridded daily in-situ precipitation observations and ERA5 reanalysis data spanning from 1961 to 2021, this study employs an unsupervised deep learning method known as the Self-Organizing Map (SOM) to objectively classify large-scale circulation patterns associated with summer persistent heavy rainfall processes in the Yangtze river basin. Four typical circulation types (P1—P4) are identified, representing distinct configurations of key circulation systems that are closely linked to the formation and distribution of anomalous precipitation. P1 and P3 represent typical single and double blocking circulation patterns at the high latitudes, respectively. In these patterns, the Western Pacific Subtropical High (WPSH) significantly intensifies and westward extends. In contrast, P2 and P4 showcase more prominent features of deepened low-pressure systems in the mid-to-high latitudes. In P2, a broad low trough extends from the lake Balkhash to the west of the lake Baikal, with a ridge to the east of the lake Baikal, forming a typical stable circulation pattern in the Meiyu season. The Yangtze region is under the influence of the low-pressure system, while the WPSH is positioned to the south. P4 shows low and high geopotential height anomalies in the west and east of the lake Balkhash, respectively, along with a cyclonic anomaly in the Yangtze region. The WPSH exhibits a noticeable northward shift in P4. P1 and P2 are supported by cold air from high latitudes, while the cold air support for P3 and P4 are weaker. Additionally, the northward advancement of the East Asian summer monsoon characterized by northward jumps of the WPSH is closely linked to the occurrence of the four circulation types and the position of rain belt. P1 and P2 mainly correspond to persistent heavy precipitation from June to early July, while P3 and P4 mainly correspond to persistent heavy precipitation from early July to August. P1 and P2 result in heavy precipitation to the south of the middle and lower reaches of the Yangtze, whereas the heaviest precipitation centers associated with P2 and P4 respectively lie in the Yangtze river and to the north of the Yangtze. Precipitation intensity corresponding to P1 and P3 also strengthens as the water vapor transport enhances. The stability of these typical circulation patterns is further analyzed. During persistent heavy rainfall processes, persistence is observed on 93.2% of all the corresponding circulation patterns. P1 and P2 typically persist for around 5 days, while P3 and P4 tend to last for approximately 3 days. The long-term trend indicates an increasing frequency of P1 and P3 occurrences, while P2 and P4 are gradually diminishing. This suggests an inclination of persistent heavy rainfall processes in the southern and middle-lower Yangtze river regions from a favorable circulation perspective. In summary, this study reveals several meaningful results, which provide a scientific guidance for clarifying the characteristics and trends of typical circulation patterns corresponding to persistent heavy rainfall.