In electronics manufacturing, manual soldering is a very important process for deciding the quality and reliability of products. Thus, this process requires skilled workers to use precise techniques. However, there are always hazards such as burns or accidents, which should be avoided to make the process safe. Ensuring the safety in manufacturing industry and plants is essential for managing the risk of accidents due to human error, especially at sites where much of the work requires manual labor. Even with the advance of automation, there are still many processes that require manual labor and they rely heavily on the skill and attention of the workers. These processes should be checked and controlled by site supervisor, but this can affect the supervisor’s own work efficiency and concentration. In this paper, we propose a skeletal estimation based motion analysis system for danger prediction during soldering process. We carried our many simulations and experiments to evaluate the proposed system. The evaluation results show that the proposed system can predict dangerous motions based on three-dimensional coordinates of a worker upper body and can be used successfully for safe soldering in factories.