Virtual Comissioning of Manufacturing System Intelligent Control

Smart manufacturing systems seek to provide adaptive intelligent strategies to manufacturing practitioners in response to environmental changes, system prognosis, and optimal solution identification. For large scale manufacturing processes, their control methodologies are expensive to train, test and develop. Virtual Commissioning, as a digital transformation method, offers a datadriven approach to automate manufacturing system knowledge so that a digital twin can be developed to visually represent manufacturing plants, numerically simulate robot behaviors, predict system faults and adaptively control manipulated variables. In this work, integrating a Machine Learning agent into the Virtual Commissioning platform Siemens Technomatix Process Simulate further expands the usage of the digital twin by training and verifying intelligent control algorithms before pushing them to the physical world for implementation. This is accomplished by transferring data between the Siemens Process Simulate Software Development Kit and Google’s TensorFlow framework to implement a Reinforcement Learning-based dynamic scheduling algorithm on a virtual manufacturing cell. For future development, control via an industrial controller will allow communication between the digital world and physical manufacturing plant so that synchronous control can be achieved. The developed control algorithms are expected to assign tasks, schedule work, generate optimal path solutions and demonstrate control robustness.