A prediction-based online soft scheduling algorithm for the real-world steelmaking-continuous casting production

Optimal scheduling of steelmaking production contributes to boosting productivity, reducing costs and achieving sustainable manufacturing for an integrated steel company. However, the optimal schedule is always difficult to implement in the real-world production system, because its optimality and feasibility are affected by various uncertain factors. In this paper, we study an uncertain scheduling problem arising from the steelmaking-continuous casting (SCC) production process which considers the cost and penalty objectives. To solve this problem, we propose a prediction-based online soft scheduling (OLSS) algorithm which belongs to predictive-reactive approach. In the proposed algorithm, a surrogate model named Gaussian process regression (GPR) is used to predict the characteristic index, slack ratio, which is able to trade off the objectives between the cost and the penalty of cast-breaks. When new batches are released to the shop floor, the soft schedule including critical decisions and characteristic indexes is determined by a dynamic optimization algorithm based on the predicted value. In the reactive phase, a heuristic method is presented to determine other non-critical decisions. Finally, the computational results show that the OLSS outperforms other algorithms in penalty objective, and obtains approximate effects in cost objective.