Solving the steel continuous casting problem using a recurrent neural network model

In this paper, we consider a multiple parallel device for the steel continuous casting problem (SCC) known as one of the hardest scheduling problems. To our knowledge, this is the first work that offers a model of artificial intelligence for the SCC, in particular a recurrent neural network (RNN) with long short-term memory (LSTM) cells that are executed in the cloud. We formulated the mathematical model and implemented a LSTM to approximately solve the problem. The neural network offers training on a big dataset of 10,000 real use cases and others generated randomly. The performances of the proposed model are very interesting, such that the success rate is 93% and able to resolve large instances while the traditional approaches are limited and fail to resolve very large instances. We analysed the results taking into account the quality of the solution and the prediction time to highlight the performance of the approach.