Virtual Commissioning and Machine Learning of a Reconfigurable Assembly System

The digital twin application in manufacturing is mainly based on the virtual simulation model of a digital twin to build a solid model, which is applied to the product processing and assembly to achieve precise production control. This paper presents a virtual commissioning digital twin model for the modularized automatic assembly system running in our lab. First, the Siemens NX MCD software tool is used to develop the virtual commissioning digital twin model for the system. Then the different working scenarios are simulated and implemented in the virtual physical simulation environment. The data from the proposed virtual commissioning digital twin model is collected and trained with 6 different machine learning algorithm such as Logistic Regression (LR), Linear Discriminant Analysis (LDA), K-Nearest Neighbors (KNN), Classification and Regression Trees (CART), Gaussian Naive Bayes (NB) and Support Vector Machines (SVM). The advantage of our newly developed virtual commissioning model is that it is able to simulate different working conditions without risk and cost-free. It is also convenient to mimic the worsening working status and failed operation scenarios which need long time to collect for the real system. We use the collected data as input for the machine learning to implement the system monitoring and predicting. The machine learning results for 6 learning algorithms are presented and it shows the possibilities and advantages of our proposed virtual commissioning digital twin model.