The workload of pilots has an important impact on the safety of aircraft operation. Predicting the workload of pilots is an important means for demonstrating the cockpit design with airworthiness regulations during aircraft development and certification. A simulated flight test is designed for a civil aircraft to collect pilot physiological data and National Aeronautics and Space Administration （NASA-TLX） evaluation values. Using pilot physiological data as the input and NASA-TLX evaluation values as the output， a pilot workload prediction model is established based on the particle swarm optimization-support vector regression （PSO-SVR） model. Compared with the support vector regression （SVR） model with default parameters， the prediction accuracy of PSO-SVR model is 7.5%， 9.5%， 7% and 5.8% higher than those of the SVR model under four different scenes. The results show that the prediction model based on PSO-SVR has higher accuracy in pilot workload prediction.