Akademik Veri Yönetim Sistemi
Wood Material Science and Engineering, 2024 (SCI-Expanded)
The assessment of the electrical properties of materials holds significant importance across various fields, from industrial heating to sensors, and knowing their complex permittivity ∈*, real (∈′), imaginary (∈′′) components, and loss tangent (∈′′/ ∈′) are the key points. The proposed study allows us to predict the complex dielectric parameters at 2.0–7.1 GHz frequency intervals of any natural wood classified as softwood based only on density information with an 89% confidence level requiring no more microwave measurement. Three natural wood classified as softwood and commonly preferred in the wood industry in Turkey were examined. A total of 16,500 first set data from s-parameter measurements, 500 raw data for each measurement, were used in the analysis, and a machine learning (ML) model was developed, which internally used a random forest algorithm for training. Initially, the model’s performance was evaluated using test data. Subsequently, the model’s prediction results were compared with the second round of measured data, which the model had not encountered before. The overall model performance on test data, as quantified by the mean absolute error, mean squared error, root mean squared error, and regression metric, is calculated to be 0.34, 0.71, 0.84, and 0.89, respectively. Results suggest that the model’s predictions closely align with the actual values, signifying enhanced model performance. Automatic classification of wood materials commonly used in industry is possible using the designed machine learning framework. In this way, the exact suitable wood material can be selected for any industrial application according to electrical parameters estimated by the model without the need for a sophisticated measurement setup.