Seneviratne’s team begins by simulating the manufacturing process before making any parts. “We can virtually simulate the process on multiple different machines, because each machine has different capabilities, and then select which machine works best for a particular part without touching any composites,” he explains.

The challenge, Seneviratne says, has been programming a machine learning algorithm that allows for defects. “It’s very hard to make the robot make mistakes. But how are you going to create a massive database of defects when the robot doesn’t want to make defects? For that we are looking at different strategies to inflate the amount of limited data that we have on defects – to expand that set of experimental data and then manipulate the data and see whether we can simulate some of those other conditions,” he says.

Scaling Up to a Smart Factory

Building upon these individual strategies is the next step toward creating a large-scale smart factory. That’s how Pactiv Evergreen, a Lake Forest, Ill.-based manufacturer of food and beverage packaging, began its smart factory transition. While the company doesn’t use composites, it works with 14 different materials, including plastic, so it can serve as a beacon for composites manufacturers.

Evergreen Pactiv set a goal of transforming its manufacturing capabilities through a factory asset intelligence (FAI) program that would improve throughput on constrained assets across the operation. While the company had previously worked through several lean-driven, kaizen events to achieve progress, the management team routinely saw its experienced production workforce return to more familiar workflows. Company leaders recognized they would need to make a fundamental process change to secure the desired step-change process improvements.

The company began its Industry 4.0 journey with a pilot project that Christine DeHaven, senior director of operations, says was selected based upon the product mix in the facility, the leadership team within the facility and the potential value of capacity the company expected to unlock within that facility. The FAI program blended Internet of Things (IoT) technology, artificial intelligence and advanced analytics solutions to collect a wide range of data that the company could use to predict material flow and track equipment maintenance needs. Among other projects, the company developed a machine learning algorithm that could predict grinder blade wear and notify production maintenance teams when to replace blades in order to reduce unexpected downtime.

The initial project taught the team the importance of determining what data was necessary to achieve desired project outcomes. “During our first launch, we connected to thousands of data points across the plant and learned that, while all data is important, not all data has the same return on investment,” DeHaven says. After completing the pilot project, the leadership team reflected on what worked and what needed to improve going forward. “We modified our launches to ensure we could define and succeed with the small, sustainable wins and then build from there,” she says.