High Altitude Table Test Data Fusion Research

Aero-engine core components of the aircraft developed engine technology is alsothe core technology of the aircraft manufacturing is a difficult point. China in this fieldhas developed rapidly, but also has made great achievements, but compared to the worldaviation power, we are still poor for some distance. Aircraft as modern society is animportant means of transport and military weapons, its role in the building of social andeconomic development and national defense force is very important. To develop areliable, secure, high-performance model aircraft must be a definite breakthroughtechnology for aircraft engines. For the development of new aircraft engines, the needfor extensive testing, in order to ensure the reliability of aircraft performance, securityand performance. Highly complex operating environment within the aircraft engine testmeasurements to the presence of a variety of errors and failures, so the test data fusionprocessing, so as to get closer to the real measurement data, and then on the engine testsystem state has a real judge, and to provide data protection for the development ofaircraft engines.The paper first reviews the main theories and methods, and research status datafusion, and several multi-sensor method an advantage, and applicable to the caseanalysis and comparison. Based on previous theories and methods, combined with thecharacteristics of the experimental data of this study, obtained based on RBF Fabricclass analysis and multi-sensor data fusion method based on the degree of confidence.And the proposed method with several existing methods on a set of measurement dataapplications matlab simulation to compare the fusion results.This paper multi-sensor (horizontal) before the measurement data of each sensordata fusion filter denoising. The first application of a single sensor measurement datatime series analysis, ARMA model data according to the timing fit, get the timingsequence state estimation model. Then the state estimation equation of state of themodel as a kalman filter, combined with the original measurement data kalman filterdenoising. Filtering denoising on a single sensor measurement data during the horizontal multi-sensor data fusion. This error is smaller closer to the true value of thesystem integration data.Finally, using a combination of the RBF network algorithm associated fault data toidentify and judge. Failure data caused by sensor failure or engine component failure.Must first determine what type of fault. If a sensor fails, it will affect the accuracy ofdata fusion, so the first exclusion. The judgment of the fault data, the first to have acertain value of the reference standard for comparison. For the development of newengine, the reference standard value for a by a large number of tests, constantly sum upthe amendment process.