Feasibility of a clouds and edge computing scheme for internal due date assignment in advanced manufacturing systems

Abstract Most of the existing edge computing applications in manufacturing systems are used for machine status detection, material transportation tracking, resource management, process control, and production planning, but are rarely used to support management activities. To fill this gap, this study designs a cloud and edge computing scheme for internal due date assignment, in which a sales representative visits a customer and negotiates with him/her the due date of a possible order. In the past, such tasks were accomplished by querying a deep learning system on the backend or cloud server, which was affected by difficulties such as cloud computing application licensing, network speed, and bandwidth. To overcome these difficulties, in the cloud and edge computing scheme, a possible order is broken down into lots that are released based on daily quotas. A deep neural network (DNN) is then constructed on the cloud server to estimate the output time of each lot, thereby obtaining the completion time of the possible order for internal due date assignment. A random forest (RF) is also built to approximate the trained DNN for internal due date assignment using edge computing. The cloud and edge computing scheme for internal due date assignment has been experimentally applied to a wafer fabrication case. According to the experimental results, the cloud and edge computing scheme provided a portable, self-contained, and instant solution for internal due date assignment that the common practice and cloud computing counterparts could not achieve. The cloud and edge computing scheme also avoided the potential leakage of confidential production condition data.