A hybrid ANN-MILP model for agile recovery production planning for PPE products under sharp demands
针对疫情等突发需求激增，提出混合人工神经网络与数学规划的两阶段方法，先预测需求峰值，再优化生产规划以最小化成本与交付时间、最大化响应能力，案例验证预测误差约1%。
Today, supply chains (SCs) have been struggling with a new type of disruption known as outbreaks such as epidemics or pandemics. This type of disruption has features such as long-term and uncertain lifespan and leads to severe fluctuations in product demand. This paper elaborates on a hybrid approach based on artificial neural networks (ANN) and mathematical programming techniques to efficiently deal with this type of disruption in the production planning of SCs. In the first phase of the hybrid approach, a multi-layer perceptron ANN model with an optimised structure is developed to efficiently predict the demand and its peak points. In the second phase, the predicted demands are considered as input in a new multiobjective agile recovery production planning model. The proposed model minimises total costs and delivery times and maximises responsiveness. A real case study in Iran is conducted to verify and validate the proposed hybrid approach. The prediction error of the ANN method is about 1 percent. According to the predicted demand, optimal decisions are determined by the proposed model. The impact of under-estimation and over-estimation of demand is evaluated in terms of total costs, delivery time, responsiveness and shortage costs in the SCs.