In June 2025, a joint innovation project was launched to explore how data-driven maintenance solutions can be implemented in traditional manufacturing environments such as sawmills – with a focus on condition monitoring and predictive maintenance in the timber industry in this particular instance. The aim is to develop and test digital tools that enable the early detection of maintenance needs in sawmill production equipment. By integrating sensor technologies and artificial intelligence, the project seeks to reduce downtime while optimising maintenance processes and improving overall production efficiency. A team of mathematicians from the POLICIES Institute is providing the data science expertise that underpins the work.
How predictive maintenance works
Predictive maintenance uses existing process and sensor data to forecast the condition of machine components through data-driven models. Ultimately, the goal is to identify potential failures early and initiate targeted maintenance before unplanned stoppages can even occur. The models involved are trained on historical data and continuously updated as new information becomes available. Here, data preparation is a key challenge: sensor data must be extracted, integrated, checked for quality, and correctly interpreted. After all, it is only through a robust, high-quality data foundation that models can deliver reliable predictions and contribute to operational efficiency gains.
Condition monitoring at the sawmill
While the underlying concept may sound straightforward, real-world implementation presents challenges of its own. For example, selecting appropriate physical parameters – such as temperature and vibration – and determining optimal measurement locations is often highly specific to the individual plant. And particularly in older facilities, critical variables and data from machinery can only be captured through retrofitting with additional sensors. The project team has addressed this by expanding the digital infrastructure with retrofitted sensors attached to key machine components (e.g. saw blades, drives, bearings, and conveyor belts). This enables real-time condition data to be collected and fed into a superordinate predictive model. As a result, data can be extracted from computer-controlled systems, linked with external algorithms, used to detect anomalies, and analysed to forecast maintenance requirements – enhancing cost-effectiveness and efficiency.
The joint project is being implemented in cooperation with DIH Süd and FH JOANNEUM .
by Renate Buchgraber
Companies involved
- Winterholz Sägewerk GmbH, Carinthia
- Bruno Ruhdorfer GmbH, Carinthia
- LSB Lärchenholz Buchhäusl GmbH, Carinthia
- Kaml & Huber Sägewerk Holzexport GmbH, Styria
