When speaking of a digital twin, we usually think of a computerized, virtual counterpart to a device or process in the real world, allowing us to study and predict its performance under varying (operating) conditions. In filtration, a classical application is the computer-aided optimization of the design of a filter medium or element to reduce time and cost spent on the building and testing of prototypes.

The relevance of such a digital twin for the prediction of the filter lifetime can be seen by the many works devoted to the simulation of flow and filtration from the microscopic length scale of the nonwoven to the macroscopic scale of the element. In the context of the fourth industrial revolution (industry 4.0), digital twins interacting with the real filter device through data exchange (will) play an important role, because they offer very interesting possibilities such as predictive maintenance, automatization of optimal operation and many more.

Obviously, a digital twin must reflect the relevant properties of the real device and therefore, one of the challenges is to identify what is relevant such that the required computational resources do not prevent a beneficial use. In this talk, we look at improving digital twins of filter elements by taking a more integrated approach with the focus on the filter media. More precisely, we consider a “chain” of digital twins, representing different stages the filter medium goes through: From “birth” (fiber laydown process), over “formation” (e.g. pleating) to “professional life” (operation in the filter element).

The material property of major interest is the (non-)uniformity of the distribution of the fiber volume fraction. The fiber laydown process determines how “cloudy” the filter nonwoven will be, which influences the filtration efficiency of the flat sheet material. The pleating creates regions of compacted nonwoven material, where the flow resistance and filtration properties can differ significantly from the flat sheet. And in contrast to a quite common assumption of computer models, filter media are not rigid bodies, but they deform under the pressure of the fluid flow. Depending on the (operating) conditions, these deformations cannot be neglected in the digital twin of the filter element.

The talk presents examples of these digital twins together with the underlying models and algorithmic aspects. We will also discuss open questions and future perspectives.