Supply Chain Optimization based on data science for spare parts of business aircraft with uncertain demand
This post-doctoral position is proposed to develop the methods by using data science and operational research approach for service level and fill rate optimizing in supply chain of DFJ, known as highly uncertain and unpredictable demand of spare parts supply chains of business aircraft.
This research will start by studying the detailed analysis of historical data, made by our previous researchers, to identify the demand patterns (smooth, intermittent, erratic and lumpy) and estimating their distributions at demand class or SKU levels.
The focus will be to extend the current research on global fill rate optimization and minimization of global investment. In this way, it is necessary to define the best fill rate for each group of products (called in DJJ as location network). It is also necessary to take into account our industrial partners’ constraints, to develop and simulate a multi-objective inventory optimization model. As presented, the main challenge concern to identify the best fill rate for each location network in order to maximize the global fill rate and minimize the global investment.
The next point concern the identification of best location after receiving ordered spear part. In fact, inventory optimization at SKU level depends to replenishment strategy and ordering policy. Knowing that the lead-time for spear part in business aircraft is long (between 3 months to 2 years), the TSL, calculated before ordering cannot be optimal after receiving the parts. That’s why, it is necessary to develop a dispatching method by taking into account the last information of consumption as well as the transportation strategy (cost vs time and uncertainties).
Finally, it is desired to develop a decision support system, easily customizable and usable for our industrial partner. The proposed tool must communicate with our in-house developed demand forecasting and inventory optimization tool. The existing platform, developed by previous researchers of this project used R and Shiny, hence, it is desirable to continue whit R.