Forecasting Aircraft Corrosion from NAVAIR Inspection and Repair Data

This paper describes the effort to develop corrosion prediction algorithms down to aircraft part level utilizing inspection and repair data found in the NAVAIR DECKPLATE, ADCS and MRP-II database systems.  The developed analysis approach is composed of four steps: 1) Censoring and Filtering of the data based on the dates of the corrosion failure observations, 2) Non-parametric Maximum Likelihood Estimate (MLE) of the data to obtain composite data "points" as a function of time using Turnbull analysis, 3) Weibull Distribution curve fitting to the MLE points to obtain a parametric for statistical forecasting,  and 4) Cox-Hazard Probability Model for multi-variant analysis of secondary data parameters such as proximity to the ocean, deployment history, etc. 

It was concluded that the existing NAVAIR inspection and repair databases do not have the geometric granularity needed to perform forecasting estimates at part level.  Steps are currently underway to add the necessary level of indenturing in specifying the location of the corrosion which will allow such forecasting in the future.