<p>The mental workload (MWL) state of operators is vital to operational efficiency and personal safety in the man–machine system of armored vehicles. In order to assess MWL in real time, fifty subjects conducted the strike task and secondary tasks based on a virtual simulation system, and the data of subjective assessment, EOG, ECG and task performance were collected. The results indicated that the task stages and secondary task types had a significant influence on mental workload (p &lt; 0.05). The mental workload at strike&amp;report stage was significantly higher than that in search phase (p &lt; 0.05), and the mental workload in visual and cognitive subtasks was significantly higher than that in the control group (p &lt; 0.05). Moreover, the significance analysis results demonstrated that the standard deviation of pupil diameter and head rotation, and mean value of eyelid opening of EOG indexes, SDNN and CV of ECG indexes, and search time and object destruction time of task performance indexes had a significant impact on mental workload (p &lt; 0.05), while heart rate showed no prominent influence on mental workload (p &gt; 0.05). A comprehensive evaluation algorithm for identifying mental workload was designed by using support vector machine (SVM) and Dempster-Shafer (DS) evidence theory as feature fusion algorithm and decision fusion algorithm respectively. The classification and prediction accuracy of proposed model was 91.9%, which was higher than that of the model using single index or SVM. This research provides an effective method for comprehensive assessment of MWL of armored vehicle operators.</p>