A new technique is proposed for the prediction and prevention of failures in sensors and avionics. The technique will be based on a seminal discovery in multivariate regression analysis that promises to yield dramatic increases in failure forecasting accuracy. Failure probabilities, rather than being estimated on the basis of one or two basic variables, such as the total number of flight hours, will be estimated on the basis of the entire life histories of items, including electrical, mechanical, and thermal stress.This will be made possible by sets of matrix equations that will analyze all the life histories simultaneously regardless of the complexity or multicollinearity of the data. The forecasting computer model and the life histories of the equipment on an aircraft will be carried aboard the aircraft on floppy disks for use every time maintenance is performed. Base and intermediate level maintenance personnel will be able to identify items that should be replaced or repaired before the next flight. The results are expected to include longer average equipment lifetime, longer periods between preventive maintenance actions, and fewer failures.
Keywords: Failures Avionics Regression Maintenance Sensors Multicollinearity Cost-Effectiveness Forecasting
