Prediction of Required Materials for Aircraft Heavy Maintenance

The aim is to provide a method of predicting the material required for heavy aircraft maintenance using mathematical prediction models. This includes the possibility of comparing the accuracy of different models. Previous studies have not applied multiple different mathematical prediction models to real operational data, i.e., data from an Approved Maintenance Organisation (AMO). To easily operate and display the results, computer software, including a Graphical User Interface (GUI), has been developed in a MATLAB environment. Real operational data was provided by AMO organisation for the years 2016 to 2020 inclusive. A total of ten mathematical prediction models were selected for the prediction calculations. These were selected based on research and previous studies on the topic of material prediction. The user of the software has the possibility to perform the prediction for all parts in the database or only for one specific part. The results are stored in the database or displayed directly in the GUI. The model accuracies for all parts are on average about 92 %. If we use cleaned data and exclude from the predictions parts that often have zero quarterly demand, we get to an average prediction accuracy of 83 %. The most accurate forecasts are made by the Exponential Smoothing with use of Moving Average Forecasting Model, Croston's Forecasting Model and Syntetos-Boylan Approximation Forecasting Method models.