A data-driven based decomposition–integration method for remanufacturing cost prediction of end-of-life products

Abstract

Remanufacturing cost prediction is conducive to visually judging the remanufacturability of end-of-life (EOL) products from economic perspective. However, due to the randomness, non-linearity of remanufacturing cost and the lack of sufficient data samples. The general method for predicting the remanufacturing cost of EOL products is very low precision. To this end, a data-driven based decomposition–integration method is proposed to predict remanufacturing cost of EOL products. The approach is based on historical remanufacturing cost data to build a model for prediction. First of all, the remanufacturing cost of individual EOL product is arranged as a time series in reprocessing order. The Improved Local Mean Decomposition (ILMD) is employed to decompose remanufacturing cost time series data into several components with smooth, periodic fluctuation and use this as input. BP neural network based on Particle Swarm Optimization (PSO-BP) algorithm is utilized to predict the cost of each component. Finally, the predicted components are added to obtain the final prediction result. To illustrate and verify the feasibility of the proposed method, the remanufacturing cost of DH220 excavator is applied as the sample data, and empirical results show that the proposed model is statistically superior to other benchmark models owing to its high prediction accuracy and less computation time. And proposed method can be utilized as an effective tool to analyze and predict remanufacturing cost of EOL products.