Aviation Flight Data Based Pilot Workload Research

To ensure aviation security,one of the most important research fields is Aviation human factors.Pilot workload is an important component of human factors in aviation.When the pilots control the aircraft and perform task,higher or lower workloads can't keep the pilot in the best flight condition and make it easy for the pilot to make a false judgment.Studying the pilot workload can help improve the design quality of the aircraft man-machine interface and reduce flight accidents.Therefore,quantitative study of workload of pilots has great practical significance in analyzing the impact of human factors on flight safety.We establish an objective evaluation method to identify pilot workload by flight parameters for making up and upgrading the original methods.Aiming at the pilot's maneuvering characteristics,this paper selects typical flight parameters with fewer parameters and the complete flight parameters with more parameters to objectively evaluate the pilot workload.We extract the mean,variance,RMS,ARMAE,WSE features of the typical and complete flight parameters for the input of deep learning features.We propose a novel feature reduction way-sparse RBM feature reduction.Finally,we apply the CS-GP classifier to evaluate the pilot workload.Compared with the Subjective evaluation method,the results validate the feasibility of the proposed method.Through the research of this subject,we can get the following conclusions,(1)ARMAE and WSE features of complete flight parameters can successfully recognize the different flight state of pilot workload.(2)Compared with the traditional feature reduction methods-PCA and Treelets,deep sparse RBM can get a better complete flight data features.(3)CS-GP classifier can efficiently build a math model with different flight parameters of pilot workload to evaluate the pilot workload,which provides a new idea for the objective evaluation of pilots.