For example, as resin transfer systems age, equipment tends to become clogged. Resin backs up, which can impact efficiency. However, cleaning equipment more frequently than required is a waste of labor and production time that can be reclaimed with better data. With analysis of data gathered by smart sensors, composites manufacturers can shut equipment down for cleaning at the first indication of slowing throughput. Librandi says this simple step can increase overall equipment effectiveness between 8% to 15%.

  • Optimized manufacturing quality –Too many manufacturers base their material curing times on trial and error or “what we’ve always done.” However, this approach can lead to under- or over-curing. It’s a problem that manufacturers tend to solve by buffering with additional material, a costly solution. By using machine learning algorithms, manufacturers can mine historic machine interactions to identify optimal cure times, among other factors. This can prevent costly scrap and rework, which leads to significant savings.

    Waruna Seneviratne, director of the Advanced Technologies Lab for Aerospace Systems (ATLAS) at Wichita State University’s National Institute for Aviation Research (NIAR), is using machine learning and artificial intelligence solutions to address the issue of part quality with an in-process inspection system.

    “With the automated fiber placement machines that we use, efficiency is low because every time we lay up something we have to stop and send a bunch of people to inspect the part. We have to do this with every layer,” Seneviratne says. Between 30% to 70% of manufacturing time may be spent inspecting a single part, he says. Then, if a defect is identified, the production team must try to manually repair the part, with potentially mixed results.

    To eliminate this downtime, ATLAS’ system will inspect the part, as well as create its digital manufacturing twin – the data-filled virtual representation of a physical product. This digital twin can help reduce the possibility of defects by pinpointing where and why they occur.

    Seneviratne says many composite components carry the potential for defects, but if they fall within an acceptable defect limit there is not necessarily a need to repair the component. However, there are many unknowns inherent within this allowable potential for defects. With a digital manufacturing twin, companies not only know exactly where those defects are but can begin to collect data that can improve manufacturing parameters and drive down these potential defects without the need for destructive testing.