<p>A dynamic movement primitives (DMP) model is a prominent intuitive method for robots to effectively acquire new skills by mimicking human-demonstrated actions. However, it lacks sufficient adaptability to dynamically avoid obstacles frequently occurring in real-world industrial scenarios. To address this challenge, this paper presents a novel segmented DMP method (namely IRG-DMP) by integrating a new segmentation strategy, an improved informed RRT* algorithm, and an enhanced genetic algorithm (GA). The major contributions of this method are as follows: (1) A segmentation strategy is incorporated into DMP to improve its adaptability for dynamic obstacle avoidance. In addition, an informed RRT* algorithm is developed to generate effective candidate segmentation points around dynamic obstacles, thereby facilitating the subsequent optimization process for identifying optimal segmentation points. (2) The enhanced GA is used to create an optimal integrated regeneration trajectory to achieve the multi-optimization targets of learning the primary features of a demonstration trajectory, realizing an efficient obstacle-avoidance capability, and achieving the trajectory smoothness. A series of validation experiments are conducted to exhibit the superior performance of the IRG-DMP method in comparison with the baseline RRT-DMP and APF-DMP methods. Case studies of using a UR5 robot for retired battery screw disassembly are carried out to showcase its applicability in industrial scenarios. Finally, discussions on segmentation point selection strategies and optimization criteria are provided to further elaborate the fine-tuning process of the proposed IRG-DMP method.</p>