Study On Dynamic Modes In Flight Simulator

With the progress and development of human science and globalization,the civil aviation has achieved rapid development in recent decades,and the amount of serving planes also presents extreme growth.To save costs in aircraft design and research,and to ensure flight safety,the demand for flight simulators is also increasing at the meantime.Compared with using real aircraft for pilot training and aircraft design,using simulator can greatly improve the efficiency and reduce the consumption.In addition,the flight simulator has the advantages of flexible installing location and environment,so pilot training,aircraft design and theory research can be conducted without the limitation of place.However,due to the complexity of the real flight,and the equipment constraints,pilots’ feeling in the simulator is not totally same with that in a real plane.The most obvious difference is the motion feeling,so whether the simulator motion is necessary to exist,whether it affects the pilot manipulation,have been debated for over half-century.In the process of people-centered simulated flight,making sure whether the pilot is influenced by the simulator motion,and if the influence exists,knowing how to cover the differences for fixed simulators,can greatly improve the efficiency of using flight simulators and make the flight data obtained from simulators more similar to that obtained from real flights.In the 21 st century,people are faced with more and more complex and huge size of data.In aviation field,as the sensor technology and computer capacity improvement,the data size is also expanding rapidly.To deal with this,data mining was proposed.Its purpose is to extract useful information from huge size and complex data,and transform them into understandable information.This study,which is based on flight data obtained from simulator,combined with the pilot cognitive process,focuses on the following three questions: Whether the pilot performences are the same using simulators of different dynamic modes? How does simulator movement affect the pilot control? How to make the flight results using fixed simulators be the same to that using dynamic sumulators? The details are as follows:(1)Focusing on the cognitive process of pilot,this study considers the important factors that affect pilot’s manipulation,analyzes the formation process of pilot skills and the effects caused by flight training.(2)Based on flight data,the research of motion influences.This containes designing flight data identification,feature extraction and selection methods,exploring the feature importance to identify whether the flight data is affected by motion.To overcome the high dimensional problem,a dimensionality reduction is studied.The research designs and executes flight simulations under different movement modes,and identifies the flight results.(3)This study analyzes the information obtained from the surrounding environment by pilot’s sensing system during tasks.According to the information,the study builds pilot models of different movement modes.Comparing the differences between models with and without motion,it finds out the influences on pilot models caused by simulator motion.(4)The study establishes the pilot-in-the-loop simulation system,proposes the compensation method for the fixed simulator,and verifies the flight data obtained from fixed simulator with compensation is similar to dynamic simulator.The following results have been achieved:(1)From the perspective of cognition,the study has analyzed the influences on pilot’s manipulation of simulator motion,and has found out the flight tasks related to dynamic according to CCAR-25.Through the ACT-R cognitive framework and experiments,it has pointed out that experienced pilots have similar driving skills.(2)This study has proposed an identification method based on affine propagation.The method is to identify whether the flight data obtained from different movement modes is the same.An improved cuckoo search has been proposed,which solves the problem of setting bias parameter and damping factor,as well as oscillation in affine propagation.The identification results show that the simulator motion has influences on flight data in the manual climbing and non-precision approach tasks,etc.(3)A feature selection and treatment method for flight data identification has been proposed for the first time.The features of polynomial fitting function,which goodness of fit is bigger than 0.6 for the first time with the polynomial degree increase are added,and Relief method is used to select characteristics.To solve the sparse problem of highdimensional flight data,an improved diffusion mapping has been presented,which keeps the data structure and reflects the data’s global and local density information.(4)Based on the characteristics of pilot’s perception information,a multi-information fusion model is proposed for the first time.Pilot models of different movement modes have been built,and the influences of different sensing information on pilot’s control have been obtained.(5)Using Visual Studio and Mathematica build the pilot-in-the-loop simulation platform.The pilot’s control compensation and perception compensation methods have been proposed to cover the differences between fixed and dynamic simulator.These compensation methods have been verified in the simulation platform.