Wood technology

With artificial intelligence to a better wood product

Nov 19, 2019 | RAINER KLOSE
Empa scientist Mark Schubert and his team would like to use the many opportunities offered by machine learning for wood technology applications. Together with Swiss Wood Solutions, Schubert will present his latest project at the Swiss Innovation Forum on 21 November in Basel.
Cannaguitar+Sonoveneer+Eiche3-web.jpg
A guitar made from modified swiss oak veneer (Sonoveneer). Image: Swiss Wood Solutions AG
Wood is a natural material that is lightweight and sustainable, with excellent physical properties, which make it an excellent choice for constructing a wide range of products with high quality requirements - for example for musical instruments and sports equipment. Unfortunately, as most natural products, wood has a very uneven material structure that extends over several length scales. Therefore, large safety margins are often required during processing, which limit the efficiency of material utilisation. With the help of science, this drawback could soon be resolved. A key technology for this is artificial intelligence.
Neuronal networks sort out flood of data

Mark Schubert works in the research department "Cellulose & Wood Materials" at Empa. In recent years, he has worked intensively on machine learning, with the goal of optimizing the functionality of wood. Schubert and his team would now like to apply this experience to other areas of wood processing. Every processing company faces the problem of interpreting and making decisions based upon the results of the receiving inspection, which provides data on the density, moisture content, fibre direction and annual ring position of the raw wood. In order to process the wood profitably, the sequential production steps such as separation, sorting and treatment must be well planned and many process parameters set correctly.

Finally, the quality control shows whether the settings used were correct and whether the processing of the wood was successful. Employees with a sense of proportion and years of experience can often help to avoid mistakes. However, there is no holistic approach that records and analyzes the raw material and process parameters that would allow product quality to be predicted in real time.

Home_Cover.jpg
This modified swiss wood can replace tropical ebony in many applications. Image: Swiss Wood Solutions AG
Premiere at start-up company Swiss Wood Solutions

That's about to change. Mark Schubert analyses the flood of data produced during wood processing using deep neural networks and incorporates them into the manufacturing process. In order to implement this innovative approach, Schubert works closely with Swiss Wood Solutions, a start-up established at Empa under the management of CEO Oliver Kläusler. The company specialises in the refinement of indigenous woods. Using a special press and know-how built up over many years, spruce or maple wood is turned into a hard, dense special material that can replace tropical ebony. 

The compacted wood species have already been successfully used for violins, guitars and clarinets, as well as for practice swords of the Asian martial art Budo. Now, the existing trial production is to be significantly expanded. Last week the company ceremoniously put its first own industrial press into operation.

Together with Swiss Wood Solutions, Mark Schubert will now expand the range of experience so that high-tech hardwood products can be manufactured with consistent quality despite the growing amount of raw material processed. The research project is funded by Innosuisse.

Mark Schubert, Empa and the Swiss Wood Solutions team will be exhibiting at the Swiss Innovation Forum on 21 November 2019 in the Congress Center Basel. 

Empa scientist Mark Schubert and his team are using the many opportunities offered by machine learning for wood technology applications. Together with Swiss Wood Solutions, Schubert develops a digital wood-selection- and processing strategy that uses artificial intelligence.


  • Print
    •  
    •  
    •  
    •