Sawmill 4.0 Customized flexible sawmill production by integrating data driven models and decisions

Purpose and goal

Purposes and sub-goals: - Create a database with high-resolution measurement data for log´s internal properties, plank´s wood properties and the customer´s quality assessment - Train prediction models for specific customer and compare volume and quality outcomes in the different production steps - Evaluate the concept of multivariate property management regarding customer satisfaction, robustness and sensitivity to variation in input quality - Test new method for image-based quality dialogue for B2B - Show possibilities for automatic traceability in the sawmill´s process

Expected results and effects

Results show that traceability and multivariate property control increase the sawmill´s volume yield and customer satisfaction. It increases by 10% when using the boardscanner and 20% if controlling the logs that are to be sawn with CT. A method for customer interviews has been tested where plank pictures are shown and the customer chooses which planks he wants to buy. The method is as good or better as common method and facilitates the quality dialogue in B2B. The results (seven articles and a doctoral thesis) show increased opportunities for improved sales order management for future.

Planned approach and implementation

Four data collections in the project have built up a database that has been used in the project. Information on 600 logs, 1200 planks and quality results for its 3600 panel boards have been used and analyzed. Discriminatory prediction models (PLS-DA) has been trained for an external customer (planery) with high raw material requirements. The models´ validity and robustness have been described and its prediction result in relation to the customer´s actual quality outcome. Prototypes for online application, variable calculation, yield and interview has been developed and tested.

External links

Länk till projektets hemsida med sammanställning av projektets publikationer