MYCOLab Ai Logo - gold and army green

Artificial Neural Network Prediction of Mechanical Properties in Mycelium-Based Biocomposites

Summary

Scientists developed an artificial intelligence model that can predict how strong and durable mushroom-based composite materials will be. These composites are made by growing mushroom mycelium (fungal threads) through wood particles and other plant materials, creating an eco-friendly alternative to synthetic materials. The AI model learns from physical measurements and can accurately predict mechanical properties, potentially reducing the need for extensive testing and helping design better sustainable materials.

Background

Mycelium-based biocomposites (MBBs) represent a sustainable alternative to synthetic composites, produced from lignocellulosic substrates bonded by fungal mycelium. Their mechanical performance depends on multiple interacting factors including substrate composition, fungal species, and processing conditions. Traditional statistical methods are limited in capturing the complex nonlinear relationships affecting MBB properties.

Objective

To develop and validate an artificial neural network (ANN) model for predicting the mechanical properties of mycelium-based biocomposites, specifically internal bonding and compressive strength, using substrate composition, fungal species, and physical properties as input variables.

Results

The ANN achieved high predictive accuracy with R² values of 0.992 for internal bonding and 0.979 for compressive strength on training data. Internal bonding was predicted more precisely than compressive strength, likely due to microstructural heterogeneities. Ganoderma sessile and Trametes versicolor exhibited the highest internal bonding, while Trametes versicolor showed substrate-dependent compressive strength variation.

Conclusion

ANNs can effectively predict mechanical properties of mycelium-based biocomposites, reducing the number of experimental tests needed for material characterization. The model demonstrates that fungal species and substrate type are the strongest predictors, and incorporating microstructural descriptors could further improve accuracy for compressive strength predictions.
  • Published in:Polymers (Basel),
  • Study Type:Experimental Study with Machine Learning Modeling,
  • Source: 10.3390/polym17182506, PMID: 41012270
Scroll to Top