Research On Civil Aviation Engine Fault Diagnosis Based On Case Reasoning

Case reasoning is to use the experience and methods when dealing with historical cases to guide the fault location of new cases.This article mainly studies case retrieval to determine similar cases.The main problems of civil aviation engine fault case retrieval are:the similarity calculation is easy to fall into the distance trap,the characteristic parameters that characterize the fault are difficult to determine,and multiple similar cases affect the accuracy of case retrieval.the specific work of this article is as follows:(1)The use of case-based reasoning technology in engine gas path fault diagnosis to reuse historical fault cases can improve engine troubleshooting efficiency.Aiming at the problem that the existing distance similarity calculation method is difficult to distinguish between cases with similar airway characteristics,this paper proposes a new metric operator to improve the Euclidean distance.This formula improves the sensitivity of the distance formula to parameter changes,and proves the distance metric For smaller changes in numerical parameters,there is a higher resolution.(2)For similar cases that are indistinguishable based on distance measurement,this paper combines a large interval nearest neighbor algorithm.This algorithm selects similar cases that are indistinguishable by setting constraints,and then constructs a penalty function to obtain a metric transformation matrix through metric learning.The characteristic parameters reflecting the fault type are mapped to a better multi-dimensional space,increasing the distance between similar cases and reducing the distance between similar cases.(3)In view of the difficulty in extracting the characteristic parameters that cause differences in engine oil consumption,the text proposes RP-CNN to classify and predict the oil consumption rate.Firstly,it is helpful for the neighborhood rough set algorithm to determine the flight phase that affects the difference in oil consumption.Then,through a one-dimensional convolution network,the flight phase parameters are reduced in dimension and the classification accuracy is calculated.