Trend Prediction Method for Aircraft Electrical Device Contact Resistance

Abstract

A method of information granule computation is employed to predict the contact resistance for a kind of aircraft electrical relays. The aim is to reduce the dimensional computation complexity and provide practical model for the evaluation process of reliability analysis and fault diagnosis. At first, numerical data sample is split into consecutive sequences by equal window length, and then the sequences are taken to construct information granules, following that, the information granules are combined together to form new time series with reduced dimensionality. Secondly, the fuzzy c-mean clustering method is used as a technique to extract the features of the data points within each information granule. The clustered information granule is defined on the basis of fuzzy linguistic terms to provide parameters for fuzzy inference system, and the fourth order information granular predictive model is constructed in the paper. The predictive model creates the relationship between past four states and future state on the basis of rule base that is produced from fuzzy c-mean clustering. Finally, the predictive model is exploited by introducing experiment data and defining different window length. The trend prediction results are shown and validated with some discussions are presented. The predicative model can reduce the dimension effectively, it is also practical and simple compared with other methods such as neural networks computation and regressive analysis, especially in the case of trend predication application that has highly dimensional measurement data.