فهرست مطالب zahra jahanilomer

In this study the GMDH neural network based on genetic algorithm to predict the physical and mechanical properties of particleboard laboratory scale has been used. predict the mechanical and physical properties of particleboard we used input parameters such as neural network including time of press closing (10،20 and 30) s، moisture mat (8،10،12 and 14) % and temperature press (150،160،170 and 180) °C as the input data and the output data is used as the physical and mechanical properties. The efficiency of these techniques evaluated with statistical criteria of mean square error (MSE)، root mean square error، (RMSE)، mean absolute deviation (MAD) and the correlation coefficient (R2). Results showed that the values of MSE، RMSE and MAD for properties MOR، IB، TS24h، TS2h، WA2h and WA24h is low. Errors were obtained for the MOE model was very high. According to the values obtained from MOE is not the appropriate model to predict. R2 values obtained from the test and training set properties for MOR، IB، MOE، TS24h، TS2h، WA2h and WA24h more than 0. 91%، Which reflects the performance of these models is better.

In the past decade، artificial neural networks have been used as a powerful tool for modeling and prediction in many scientific fields. In this study، the feed-forward multilayer Perceptron (MLP) was utilized and trained by back propagation (BP) algorithm with Levenberg-Marquardt numerical optimization technique via Matlab software. Temperature of press (°C)، mat moisture content (%) and press closing time (sec) were used as inputs، Water absorption (WA2، 24h)، thickness swelling (TS2، 24h) and density were the outputs of neural network model. This technique will increase network versatility and decreases the effect of undesirable and weak data. The modeling and prediction was done based experimental data and the forecasting results were compared with real data. The efficiency of these techniques evaluated with statistical criteria of mean square error (MSE)، root mean square error، (RMSE) and the correlation coefficient (R2). The results showed this ANN model could accurately describe the water absorption، thickness swelling after immersion for 2 and 24 hours، and density of particleboard