Overview:

This project deals with “Aircraft components reliability”, which is a key concept in the aeronautical industry. The main purpose of this project is to utilize some statistical models in order to predict future failure of aircraft parts, and then to optimize the management of spare parts stock.

This project is carried out in favour of Airbus Flight Hour Services (FHS), an Airbus department that provides stocking and repairing services to airline companies all around the world to minimize the time and cost of replacement of aircraft failed components. At airlines, inoperative units on airplanes are removed and then are replaced with serviceable units that are provided from their on-site stock. Airbus FHS here plays two important roles: providing serviceable units for the site stock and repairing unserviceable units at repair stations (the OEMs).

The main objectives of the project is to:

• define component lifetime forecast models and then by using these models,

• determine if the current stock of spare parts is enough to full fill all the needs until the end of the contracts. If not, determine when the stock will be empty and will thus require to be replenished with new or repaired parts.

In the new situation due to the pandemic, the airlines have less aircraft flying and more contracts ending. Consequently, FHS need to fine-tune strategy “repair or stock unserviceable” to optimize the stock management. So the failure forecast of aircraft parts is a significant underlying contributing factor in the proposals of such strategy.

Tutorial:

We first download the project and then, the lifelines library in Python is required to execute the codes of project:

pip install lifelines

The version number of each package is presented in requirements.txt.

The notebook file /doc/FlightDataAnalysis.ipynb includes all of codes for the simulation process as well as the graphs to visualize.

The input.txt file is to calibrate the parameters of simulations:

• <Data file name>: name of data file which must be string.

• <Confidence level>: confidence level coefficient which must be in (0,1).

• <Company name>: company that we want to forecast, if company==0 so we want to forecast for all companies. This parameter is only useful for Number_Failures_Forecasting() function.

• <Unit type>: part number we want to forecast for.

• <Month> and <Year>: the moment that we want to forecast in the future. This is not useful for Time_Forecasting() function.

• <Number of type unit actually on stock>: number of product of part number <Unit type> available on stock at the moment. This only useful for Time_Forecasting() function.

• <Service level>: percentage to which customers' need should be met. Only useful for Time_Forecasting() function.

• <Repair rate>: percentage to which failed products are replaced by repaired products instead of replaced by new.

Attention: The data set format (clone name, sheet name, etc.) should be similar to the format of the data on /DataSet/ and the data file should be named without spaces.

For estimating of the number of failures at a determined time in the future, we run Number_Failures_Forecasting.py file with the command:

python Number_Failures_Forecasting.py

For evaluating the failure number as a function of time, we run Evolution_Number_Failures_Forecasting.py file:

python Evolution_Number_Failures_Forecasting.py

For estimating of the moment that we can not immediately afford the need of customers (called out-of-stock date), we run Time_Forecasting.py file:

python Time_Forecasting.py

Finally, the /output directory contains simulation output results, which will be generated automatically when running our code.