Research On Parallel Computing For Takeoff Performance Analysis Based On OpenMP

As the domestic large aircraft industry is booming,the performance computing software that supports domestic civil aircraft will also face new challenges.Although it can meet the basic needs of daily airline operations,the increase in usage has shifted users’focus to improving the calculation speed of the software without sacrificing accuracy,benchmarking the calculation efficiency of performance computing software for Airbus and Boeing models.The takeoff phase is a frequent phase of flight accidents,and airlines attach great importance to accurate and detailed pre-flight performance analysis before takeoff.The takeoff performance analysis module has become a heavily used function of performance software.The ultimate goal of takeoff performance analysis is to generate takeoff analysis tables for use by operations control and crew preparation before release,involving performance parameters such as maximum takeoff weight and takeoff speed.When there are many factors affecting the airport conditions,atmospheric conditions,etc.,the calculation amount is huge,and there are many levels of iterative cycles.In order to ensure the accuracy and reliability of the calculation results,the software program is based on the first law to establish a physical model,and the calculation process requires the use of a large number of takeoff speed,distance,and gradient calculation models,which involves complex logic.As aircraft manufacturers continue to optimize the design of aircraft models in later stages,the difficulty of simulation calculations will also increase,making it more difficult to meet the ideal calculation efficiency of airline users.This article describes the mathematical models for calculating the takeoff speed,takeoff distance,and takeoff weight limit involved in takeoff performance analysis using the integration method,and elaborates on the usage of the OpenMP parallel specification in detail,establishing the theoretical basis for the parallel construction of the takeoff performance analysis algorithm.By studying the maximum takeoff weight calculation model of the aircraft,the entire takeoff performance program is analyzed,and the calculation models of the hot subprograms are established.First,the AFM module and TLP module of the performance software are designed with loop parallelism.The TLP module includes the outermost maximum takeoff weight calculation and the bottom-level balance V₁ loop parallel algorithm design logic.Secondly,the middle layer of the calculation model for the calculation of the takeoff point of the TLP module’s maximum takeoff weight is designed with a task parallel algorithm.Finally,a nested parallel design scheme is proposed,and the parallel models involved are programmed and implemented using Fortran language based on OpenMP,and optimization results are output.After completing the programming implementation,the serial program calculation time required for the AFM module’s loop parallel algorithm design and the parallel program calculation time（including 2,3,and 4 thread parallelism）were tested and analyzed,and the conclusion was drawn that the parallel algorithm can greatly shorten the running time of takeoff performance analysis calculation.When using four threads for calculation,the program running speed can be 3.15 times faster,This fully demonstrates the computational speed advantage of parallel algorithms in solving performance problems.And analyze the time required for parallel algorithms with different thread numbers,and obtain the conclusion that the existing takeoff performance model’s parallel computing acceleration ratio increases with the increase of thread numbers,while the parallel efficiency decreases with the increase of thread numbers.The TLP module uses 2 threads for loop design parallel performance analysis,with an average parallel acceleration ratio of 1.72 times and a maximum of 1.89 times;The average parallel efficiency is 86%,with a maximum of 94.64%.For the design of Task parallelism algorithm,first use a single thread to perform single thread operation on the first,second and last sections of the channel,and then use four threads to run the remaining weight restriction.The calculation time after parallel is the sum of the longest thread in the subsequent four threads plus the weight restriction time of the channel section.However,due to the complex calculation conditions,the length of weight restriction calculation time varies in different cases,Therefore,the acceleration ratio of Task parallelism is uncertain,but it can still effectively shorten the calculation time of maximum takeoff weight.The results indicate that using OpenMP parallel processing technology to optimize the computing speed of ARJ21 performance software is effective,and it effectively saves the computational time required for takeoff performance analysis.