Digital Twin-Driven Predictive Quality Analytics

Abstract

In the digital environment, sectors such as manufacturing and industry have the highest demand for intelligent quality monitoring systems for supporting the quick and accurate prediction and enhancement of quality. Current quality control techniques are mainly based on phase-wise inspection, back-analysis etc., which are surely not enough and not even adequate to industrial quality control processes of dynamic and data-rich organizations of processes. As an example, in this work, a Digital Twin based Predictive Quality Analytics framework is presented, consisting of the combination of real-time sensor values, Industrial Internet of Things (IIoT) devices and cloud computing as well as Machine Learning algorithms for pro-active quality management. The new approach involves making a virtual copy of the real production system, and constantly synchronizing operation data between physical and virtual systems. Deviation detection based on ATV models (refer to norm values); Deep learning and anomaly detection can be applied to identify deviations, potential quality issues and suggestions for appropriate corrective action. It consists of five key components: data acquisition layer, digital twin modelling layer, cloud based analytics layer, predictive intelligence engine and the decision support dashboard. The experiments show how the proposed system will have an immense impact on defect prediction accuracy, production down time, material wastage and reliability. Moreover, with real-time feedback control, the well adaptable optimization and learning in an industrial context is possible. Based on the ultrasound and the calculated devol(std) value, the study shows how quality assurance applications can be realized to enable next generation manufacturing ecosystems via the support of different Industry 4.0 applications and support of the digital twin and AI.