Alpine inspiration

Wood can be split parallel to the fibers with minimal energy and almost no losses. Image: Empa

Traditionally, shingles are split by hand from log segments, while industrial processes rely on pneumatic splitting tools. “Shingle production shows us how wood can be processed in an energy-efficient and material-efficient way,” explains Burgert. “Wood can be split parallel to the fibers with minimal energy and almost no losses.” This chipless wood processing considerably increases the sawn timber yield, which, at around 60%, is significantly lower in Swiss sawmills.

Cross-section of the newly developed wooden panel made from split sticks: The stabilizing fiber orientation and high material yield make it a potential alternative in timber construction. Image: Empa

In traditional shingle production, however, only selected high-quality softwood is usually used. In order to adapt the process for lower-quality hardwood species and to split longer sticks, the researchers rely on a two-stage splitting process. First, flat elements are separated, which are then further processed into wooden sticks of the desired dimensions. On a laboratory scale, the researchers adapted an apparatus for splitting firewood for this purpose. Thanks to a multi-bladed splitting head, several boards or sticks can be produced at the same time during one splitting process.

Illustration: Empa

The splitting process produces wooden sticks in the direction of the grain without cutting the stiff and strong fibers. However, the irregular shape of the sticks poses a challenge. To overcome this, Burgert and his team are relying on artificial intelligence (AI). An automated camera system captures high-resolution images of each wooden bar, which are fed into a neural network. “With AI, we can determine important wood properties such as stiffness for each stick, regardless of shape, size or type of wood,” explains Empa researcher Mark Schubert. “If we use different types of wood of different qualities in the future, wood sorting will play a crucial role. With our machine learning algorithms, we therefore generate as much data as possible about each individual piece of wood in order to use it optimally for wood-based materials with defined properties.”

The team has pressed the first panels without sorting the wooden sticks beforehand. Even so, the potential of the manufactured demonstrators is already apparent: The panels can be produced in a highly resource-efficient manner and have mechanical properties that make them ideal for load-bearing components in the future. Despite challenges in terms of production processes, bonding, scalability and the predictability of material properties, Burgert is optimistic: “Our process has the potential to offer a sustainable alternative for the use of wood in times of accelerating climate change.”