Active vibration control (AVC) systems are essential for reducing unwanted vibrations in different dynamic systems, but researchers face several challenges during their development. These challenges include system uncertainties, actuator limitations, model accuracy constraints, etc. In this paper, we propose a novel approach to active vibration control using Reinforcement Learning (RL) to address these issues effectively. This study examines the ability of an RL-based controller to regulate dynamic systems across both high and low natural frequencies under random initial displacement conditions, and quantifies the range where the trained Deep Deterministic Policy Gradient (DDPG) agent performs effectively. To achieve this, we employ a DDPG-based RL algorithm, which is well-suited for managing continuous action spaces. The simulation results demonstrate that the RL-based vibration controller significantly improves robustness and vibration attenuation for higher and lower natural frequency systems. This work highlights the potential of using RL techniques to enhance active vibration control within a measurable performance range.