When assessing function hazard on time and function continuously varying systems, it is often difficult for the assessor to evaluate the effect of the function performance. The method based on SAE ARP 4761 has not provide an assessable way to evaluate continuous systems, but only addresses a limited number of failure modes. Therefore, we combine assessable function failure data with fuzzy set to establish membership functions. By introducing function performance and time into two-dimensional normal distribution function and its variants, we can divide function hazard into different levels. Taking the impact of UAS moto as an example, we apply the function hazard assessments based on fuzzy set and experience for a comparison. The results shows that the proposed models can be applied to the function hazard assessment.