This work combines virtual reality with emotional interaction design to explore the applications on aerospace assembly training and to improve training efficiency and improve training experience. By investigating the research on multimodal emotional interaction, the emotional interaction model in aerospace virtual assembly training is built by using emotion recognition based on physiological signals, speech interaction, highlights and videos. Taking the virtual assembly training of aerospace ball valve as an example, an aerospace virtual assembly training system with emotional interaction is constructed. Real-time emotional monitoring is performed by using physiological data collected from GSR sensors worn on the user’s finger. Three interactive schemes are provided according to different emotions:speech interaction, highlights and video. The effectiveness of emotional interaction in virtual assembly training system is verified by the efficiency of completing assembly tasks in the virtual environment. The experimental results show that the emotional interaction design can effectively shorten the training time and greatly improve the training experience of users. This work can promote the research on the multimodal emotional interaction paradigm by combining physiological signals, speech interaction, visual interaction, etc. This work verifies the effectiveness of the emotional interaction in the application of aerospace virtual assembly training, which has wide academic meaning and application prospects.