The manufacturing industry is rapidly evolving toward higher efficiency, lower cost, and greater flexibility. The production of workpieces is typically characterized by small batch sizes, diverse product types, and complex processing procedures. In practical production environments where machine degradation and random failures frequently occur, traditional deterministic flexible job shop scheduling models are often insufficient to accurately describe real manufacturing processes. To address this issue, this paper focuses on the stochastic flexible job shop scheduling problem (SFJSP) considering machine degradation and random failures. The processing equipment in a flexible job shop is further modeled as a serial production unit composed of degrading machines, thereby constructing a scheduling model that more closely reflects practical production scenarios. On this basis, to overcome the high computational cost of traditional metaheuristic algorithms in complex and uncertain environments, an intelligent scheduling method based on the D3QN (Dueling Double Deep Q-Network) reinforcement learning framework is proposed. The proposed approach simultaneously optimizes multiple performance objectives, including makespan and total production cost. Numerical experiments demonstrate that the proposed method outperforms the compared algorithms in terms of both solution quality and computational efficiency, and can effectively improve the operational performance of complex production systems.