An MBD-driven order remaining completion time prediction method based on SSA-BiLSTM in the IoT-enabled manufacturing workshop
针对物联网制造车间中多源异构数据处理难的问题，提出一种基于樽海鞘群算法和双向长短期记忆网络的制造大数据驱动方法，用于实时预测订单剩余完工时间，帮助管理者掌握生产进度波动。
The transformation of production mode leads to the need to strictly ensure the order in-time delivery. The global control ability of manufacturing workshops has become a necessary capability for enterprises. The real-time accurate Order Remaining Completion Time (ORCT) prediction allows managers to master the production schedule fluctuation, which provides criteria for workshop management. Due to the widespread deployment of Internet of Things (IoT) devices in the workshop, the difficulty of systematic processing and analysis of multi-dimensional heterogeneous data has been the main pain point of ORCT prediction. To tackle these problems, a Manufacturing Big Data (MBD)-driven ORCT prediction method based on Salp Swarm Algorithm (SSA) and Bidirectional Long Short-Term Memory (BiLSTM) is proposed. An attribute selection algorithm combined with max-relevance and min-redundancy and regularisation is used to mine key data from MBD. An SSA-BiLSTM model is proposed to achieve efficient and accurate ORCT prediction. The prediction model is supported by BiLSTM, and the hyperparameters are optimised by SSA. Finally, a case study about ORCT prediction in the IoT-enabled manufacturing workshop is presented. The result verifies that the proposed ORCT prediction method has obvious advantages over the other three traditional methods in accuracy and efficiency.