Sapa Heat Transfer AB (SHT AB) is one of three components in the corporate group Sapa Group, owned by the Norwegian company Orkla ASA. Sapa Heat Transfer AB has production in Sweden (Finspång) and China (Shanghai). The company is a world leading supplier and developer of aluminum strips used in various types of heat exchangers, primarily in the automotive industry.
In several stages of the production, which includes remelting, hot and cold milling and cutting, a part of the processed material becomes scrap metal. This scrap metal is collected and reused as raw material in the remelting factory. SHT AB currently has no automated IT support to forecasting this scrap metal; these calculations are, in the current situation, made with the help of Excel. In the calculation of scrap metal the company uses a standard value, corresponding to the average yield for all alloy combinations in the production. As a mean value is being used for all alloy combinations, the forecast becomes misleading for some alloys. The yield varies widely depending on the alloy, which means there are alloy combinations that are above the average yield and some that are below. In the current situation SHT AB has no link between actual orders and the short-term forecast of scrap metal.
The purpose of this thesis is to develop proposals on how Sapa Heat Transfer AB can and should develop its forecasting of scrap metal. This will help them getting a better accuracy in their forecasting and thereby improving the raw material planning.
In order to improve the long-term forecast of scrap metal the decomposition of the forecast for the alloy combination had to be analyzed. An improvement of the decomposition will mean an improvement of the precision for the long-term prognosis of scrap metal. A change in the decomposition will also influence the capacity planning in the remelting factory.
The action proposal for the long-term forecasting will result in a better accuracy of the forecast by replacing the mean value for the yield by a specific forecasted yield of each alloy combination. The data needed to carry through such a forecast is available in SHT AB´s databases, and by using the forecasting method exponential smoothing the yield can be forecasted. By using this method a forecast will be obtained that, with a low value of α, will not take such great impact from extreme values that sometimes occur.
The current short-term scrap metal forecast, based on a decomposition of the alloy combination forecast, can be improved by a link to the actual backlog. If this link is established the scrap metal forecast will be based on known information instead of forecasted data. A link to the backlog can only be used for a short-term forecast due to the length of the awaiting orders.
The action proposal to improve the short-term forecast assumes that a standardized forecast is developed for the scrap groups to be projected. The forecast should use an individually forecasted scrap factor produced with exponential smoothing for each scrap group. The scrap factor is based on historical data for the material planned to go into the process and the weighed scrap delivered to the remelting factory. With this scrap factor, data on how much material that is planned into the process the coming weeks and with consideration of scrap falling throughout the lead time the expected amount of scrap for the coming week / weeks can be calculated.
To get a better precision compared to the standardized forecasts, the forecasts of the most interesting scrap groups should be developed and adapted. They can be given individual values of the smoothing constant α for the prediction of the scrap factor and the way the lead time is being considered can be adjusted. In the proposal the simplification that scrap is expected to fall symmetrically distributed over the lead time has been done. In reality this is not the case. Therefore, the forecast should also be developed to take into account that, for example, more scrap is falling in the end of the lead time.