Predictive Performance of Artificial Neural Network and Multiple Linear Regression Models in Predicting Adhesive Bonding Strength of Wood

Bardak S., Tiryaki S., Bardak T., Aydin A.

STRENGTH OF MATERIALS, cilt.48, sa.6, ss.811-824, 2016 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 48 Sayı: 6
  • Basım Tarihi: 2016
  • Doi Numarası: 10.1007/s11223-017-9828-x
  • Dergi Adı: STRENGTH OF MATERIALS
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Sayfa Sayıları: ss.811-824
  • Anahtar Kelimeler: artificial neural network, prediction, bonding strength, multiple linear regression, model comparison, HEAT-TREATED WOODS, MOISTURE-CONTENT, SHEAR-STRENGTH, TIME-SERIES, ELASTICITY, FORECAST, MODULUS, MOE
  • Karadeniz Teknik Üniversitesi Adresli: Evet

Özet

The purpose of this study was to develop artificial neural network (ANN) and multiple linear regression (MLR) models that are capable of predicting the bonding strength of wood based on moisture content, open assembly time and closed assembly time of the joints prior to pressing process. For this purpose, the experimental studies were conducted and the models based on the experimental results were set up. As a result of the experiments conducted, it was observed that bonding strength first increased and then decreased with increasing the wood moisture content and adhesive open assembly time. In addition, the increased closed assembly time caused a decrease in bonding strength of wood. The ANN results were compared with the results obtained from the MLR model to evaluate the models' predictive performance. It was found that the ANN model with the R (2) value of 97.7% and the mean absolute percentage error of 3.587% in test phase exhibits higher prediction accuracy than the MLR model. The comparison results confirm the feasibility of ANN model in terms of predictive performance. Consequently, it can be said that ANN is an effective tool in predicting wood bonding strength, and quite useful instead of costly and time-consuming experimental investigations.