Generic framework for developing process digital twin applicable to high value-added manufacturing

Abstract

Our team has developed a generic modular framework to create Process Digital Twin (PDT) that can be customized for any process. Elements of this framework are (i) configuration of manufacturing process inputs as seen by user, (ii) capturing variation of manufacturing inputs (iii) build accurate and fast multi-parametric numerical solution using advanced model order reduction (MOR) techniques, (iv) Extract features for quality (FFQ) using specialized algorithms and (v) predict probability of success from quality models and (vi) capability to receive post-production process and quality data and use machine learning algorithms to automatically update various models. Irrespective of the physical and geometrical complexity of individual process, our numerical techniques assure a real time (< 10 sec) multi-parametric solution. Primary output of process digital twin (PDT) is predicting probability of success of a process based on all manufacturing inputs both material and physical. This allows us to use PDT to maximize production success by adjusting controllable inputs apriori. We have successfully used the framework to build digital twins for complex composite manufacturing processes that produce fan blades for aircraft engines and implemented it into manufacturing process. We will present some examples to demonstrate the paradigm shift of reactive to adaptive manufacturing.