Predicting Wood Thermal Conductivity Using Artificial Neural Networks

Stavros Avramidis, Lazaros Iliadis


An artificial neural network model that estimates wood thermal conductivity under a wide range of conditions of moisture content, temperature and apparent density was developed and tested with literature-obtained experimental data. The optimal network was determined to consist of an input layer, three hidden layers, and one output layer following the feed forward network structure and more specifically the back-propagation algorithm. Each of the three hidden layers of the ANN consisted of eleven neurons. The Neuralworks software package was used for the determination of the network structure and architecture, and for the training and testing phase. The evaluation produced an R2 value equal to 0.9994 and a RMS Error equal to 0.0123, thus proving that the developed ANN model is a reliable approach with powerful predictive capacity towards the estimation of thermal conductivity and it can be used by researchers under a wide range of conditions.


Artificial neural networks;density;moisture content;temperature;thermal conductivity coefficient;wood

Full Text:



  • There are currently no refbacks.