Pipelines are vital for transporting oil and natural gas, which are always facing many security challenges during the transportation process. Pipeline failures can lead to significant economic losses and injuries, therefore risk assessment and maintenance are crucial. To address the lack of precise data and inherent uncertainty in evaluating pipeline risk, this paper proposes a new approach for assessing pipeline system reliability. This method integrates the weakest t-norm algorithm, fuzzy fault tree analysis, and Bayesian networks to compute the reliability of natural gas pipeline systems, mitigating the issues of fuzzy accumulation and bias found in traditional fault trees. Furthermore, research is conducted on optimizing maintenance strategies under hybrid maintenance. The maintenance optimization model fully considers the specific requirements of pipeline reliability in different risk areas. The proposed method can provide corresponding optimal maintenance solutions for different risk areas, which can better meet the actual needs of engineering.