This study advances flight control systems by integrating deep reinforcement learning to enhance fault tolerance in fixed-wing aircraft. We assess the efficiency of Cross-Entropy Method Reinforcement Learning (CEM-RL) and Proximal Policy Optimization (PPO) algorithms in developing an adaptive stable attitude controller. Our proposed frameworks, focusing on smooth actuator control, showcase improved robustness across standard and fault-induced scenarios. The algorithms demonstrate unique traits in terms of trade-offs between trajectory tracking and control smoothness. Our approach that results in state-of-the-art performance with respect to benchmarks, presents a leap forward in autonomous aviation safety.