Using Neural Networks to Solve a Disassembly-to-Order Problem

Neural Networks (NN) technique is widely used to solve problems with complex or un- known input-output relationships. In this paper, NN concept is implemented in order to solve the disassembly-to-order (DTO) problem. DTO is a system where a variety of returned products are disassembled to fulfill the demand for specified numbers of components and materials. The main objective is to determine the optimal number of take-back EOL products for the DTO system which satisfy the desired criteria of the system. Since take-back EOL products are in uncertain conditions, model formulation is challenging. NN, which is capable of recognizing the hidden relationship or pattern of a given input-output data, is a very promising technique to solve the DTO problem. In this paper, we use NN to solve the DTO problem. First, NN is trained by a set of data which has the components demanded as input and optimal number of take-back products as output until the rela- tionships between the two are recognized. After that, the trained NN is used to obtain the optimal number of take-back products for the component demands with unknown solutions. A numerical example is considered to illustrate the methodology.