Increased Recycled Aluminum Content during Remelting by Incorporating Compositional Uncertainty

"Compositional uncertainty in secondary materials and variable processing conditions inhibits secondary material utilization in the production of aluminum alloys. Efforts to implement optimization models that direct batch planning have limited industrial presence due to factors such as the insensitivity of these optimization models to dynamic plant circumstances. This paper describes a chance constrained model formulation that generates lower cost, robust batch plans with significant flexibility towards changes in processing conditions. The principal attribute of the chance constrained formulation is its inherent mitigation of risk by diversifying the batch plan. This paper compares aluminum remelter batch planning using the chance constrained formulation with a traditional deterministic model and a manual batch planner. With the expectation that volatile material prices, restricted scrap material availability, and pressure by consumers and legislation for increased recycled raw material input will continue in the future, batch planning tools that inherently diversify a production portfolio will become increasingly advantageous for industrial aluminum remelters.IntroductionAluminum remelters must develop strategies to incorporate lower cost secondary materials into finished alloys to remain competitive amidst increasingly demanding market and regulatory conditions. Determining the optimal combination of incoming raw materials to produce an alloy that meets compositional specifications presents remelters with a complex problem that has driven the development of computational tools commonly referred to as batch planning or blending algorithms. The ubiquity of the ingredient or raw material blending problem in production has motivated ongoing development of batch planning algorithms since the classic nut-mix problem in the 1950's [1]. The most common model formulation for industrial aluminum remelters are deterministic, linear optimization programs [2]. These blending problems determine the optimal combination of raw materials by minimizing cost subject to constraints around demand, availability, and compositional specification. Despite extensive research and moderate computational requirements, batch planning algorithms are inconsistently used in industry. Insufficient incorporation of the complexity of real world remelting conditions into these models provides an obstacle for their implementation [3]. For example, many batch planning algorithms do not explicitly incorporate the impact of compositional variation of the raw materials on overall batch cost. Additionally, they do not allow for endogenous specification of tolerance to error in batch planning, so provide little flexibility where a batch planner would like to incorporate materials of greater uncertainty. Incomplete accounting for quality variability undervalues secondary materials leading to overly conservative use [ 4, 5]."