Validation of a digital twin: Part distortion in heat treatment

Abstract

This paper presents a study on multi-process manufacturing optimisation, employing a sophisticated data analytic model based on Gaussian Processes. Through strategically designed aluminium trials and rigorous model evaluation using leave-one-out cross-validation, the efficacy of the model in identifying optimal parameters is demonstrated. An interactive dashboard, powered by the Anaconda Panel(c) library, enables real-time insights into the impact of varying parameters on outcomes. Experimental validation showcases the model’s ability to reduce distortions yet highlights the dynamic nature of manufacturing processes. The iterative refinement of optimal parameters based on real-world observations underscores the adaptability of the model. This study emphasises the importance of combining advanced data analytics with practical experimentation for enhanced precision in modern manufacturing, paving the way for further advancements in the field. Real-world validation of predicted optimal parameters reveals unexpected distortions, highlighting the dynamic nature of manufacturing processes and the necessity for continuous refinement. Further analysis uncovers a discrepancy in distortion outcomes, emphasising the need for vigilant investigation and iterative refinement. By incorporating the latest experimental findings, the data analytic model has generated updated optimal parameters, effectively minimising part distortion. This study underscores the significance of data-driven decision-making and the continuous integration of new experimental results in the pursuit of optimal manufacturing parameters, paving the way for further advancements in manufacturing optimisation.