Accelerating Design Processes Using Data-Driven Models

Abstract

Data-driven modeling is a new paradigm in science and engineering for dealing with design problems. In the past, collection of data was problematic and scarce, and was mainly done to document experiments and verify research hypotheses. Today, we are facing information explosion, as data became affordable and simple to gather, store, and distribute. This became possible due to availability and prevalence of sensors, high-volume storages, cloud services, and internet of things solutions. Abundant data accessibility causes serious challenges to the modern engineering, and we should learn how to take advantageous of them in design processes. Data-driven models have many important applications. Complete models can be created based on gathered information without knowing the underlying physics. Existing, working models can be improved, verified, and extend on account of the data. In many important applications where strict simulation times are crucial, like in control and human in the loop systems, data-driven modeling can be an instant solution by providing highly efficient models. Moreover, the new paradigm allows for emergence of novel engineering fields like digital twins and virtual sensors. It can be stated that there are no modern design processes without data utilization and, consequently, data-driven modeling.