The world is rapidly moving towards a digital age, and industries are following suit. The combination of advancements in sensors, communication, and big data has fuelled the growth of industrial digitization. However, the true breakthrough in this process has been the success of artificial intelligence (AI) and machine learning (ML). Particularly, the development of deep learning has been a significant milestone in the journey from digitization to digitalization. Deep learning allows for the modeling of highly complex problems, without the need for specific algorithms, by leveraging vast amounts of data. Nonetheless, this approach is extremely data-intensive and computationally demanding, limiting its successful application to areas where abundant data is available, such as image classification and natural language processing. The creation of a fully connected deep network solely from experimental data for engineering remains a distant goal. Consequently, a hybrid approach that combines engineering knowledge with machine learning has emerged. These techniques essentially incorporate human expertise and computational physics simulations into machine learning models, thus utilizing underlying knowledge and decreasing the dependency on data.

This presentation delves into ongoing research conducted by the author's laboratory, focusing on applications in industrial settings. These engineering and data-driven algorithms, capable of performing predictive and diagnostic tasks, are integrated into the concept of a digital twin.