Abstract: Nuclear safety serves as the foundational prerequisite for the sustainable development of nuclear power. Statistical analyses reveal that human factor errors areone of the predominant contributor among various elements in nuclear power plant accidents. As a pivotal domain within human factors engineering research, personnel performance evaluation represents a critical safemechanism for nuclear safety. Conventional performance assessment methodologies predominantly rely on post-incident static evaluation, which exhibits inherent limitations including excessive subjectivity and unidimensional assessment criteria, rendering them inadequate for accurate performance quantification in dynamic human-machine interaction scenarios. This study proposes an integrated performance evaluation framework incorporating behavioral, psychological, and physiological multimodal data fusion. The proposed framework implements a dynamic weight adaptation mechanism coupled with a fuzzy synthetic evaluation model, effectively addressing multidimensional data integration challenges. Experimental validation conducted through the PCTRAN simulation platform demonstrates the method's capability to autonomously adapt to task complexity and stress conditions, enabling dynamic and quantitative performance assessment. The results confirm superior measurement accuracy in complex dynamic operational environment.