Research On Performance Analysis Technology Of Aero-Piston Engine Based On Time Series Data Prediction

Changes in the performance of aviation piston engines have the most direct impact on the operational reliability of aircraft.How to use good methods to identify problems based on their performance variation rules,provide reliable basis for condition based maintenance,and thereby improve the stability and reliability of aircraft operation is a very practical and meaningful work.This article uses a data based analysis method to establish a data prediction model for analyzing the performance changes of the engine,and studies the data analysis results and possible problems with the engine,assisting aircraft maintenance workers in aircraft maintenance work,improving maintenance efficiency and reliability.The research content will be divided into two parts: data analysis visualization and prediction research based on flight time series data.Perform correlation analysis on parameters that affect engine performance,screen out data that reflect engine performance,and use it for engine performance trend analysis.Establish an artificial neural network time series data prediction model.Evaluate engine performance changes through analysis and comparison of predicted values and actual values.Finally,further analyze the data to present detection and performance analysis results.The main work content is arranged as follows:(1)Visual analysis and parameter screening.When an aircraft is flying,engine performance parameters at different times,at different flight altitudes,and under different flight conditions will present different results,and these data will have a certain trend of change over time under the same operating conditions.However,there are some problems in the application research of QAR timing data recorded by aircraft in actual operation,such as weak regularity of data recording,missing values,and data noise.This requires a large amount of visual observation and data cleaning work,so that the actual generated data can be applied to experimental research.(2)Artificial neural network prediction for engine performance analysis research.The specific research takes part of the flight data recorded by the G1000 system of the Cessna 172 R aircraft as the research object,and analyzes the parameter data recorded by the flight data under the condition of time series.Reflect engine performance trends with engine exhaust temperature changes,and predict and analyze engine performance trend data.Use multiple models to predict exhaust temperature,construct filtered exhaust temperature related parameters for artificial neural network prediction,compare their prediction performance with the advantages and disadvantages of regression model prediction algorithms through experiments,and explore the performance changes of the engine based on the change trend of exhaust temperature over a long period of time.By predicting exhaust temperature,explore possible problematic nodes over a period of time,establish a time series engine performance analysis model,and use injected fault sample data to test the analysis and prediction model.