Redes neuronales en la predicción de propiedades de láminas para tableros de madera contrachapada durante el proceso de secado en Arboriente S.A.

Abstract

To avoid delays in production, caused by manual measurement of raw material properties, this research is presented, the objective of which is to predict the output humidity and the waviness of the sheets for plywood boards, during the drying process in ARBORIENTE SA. First, a network of ten thermocouples was installed that monitor the temperatures. Then, Doncel, Tamburo and Zapote were selected as the wood to be studied. Controllable factors were thickness, speed, and type of wood; the non-controllable inlet temperature and humidity; having as variables the output humidity and the percentage of undulation. Finally, 54 experimental runs were carried out, determining that the output moisture content and the waviness depend to a great extent on the speed of the rollers since this controls the time that the sheets remain inside the dryer, while the thickness is not. statistically significant for the prediction of the output variables; with the data obtained, the prediction of the response variables, between a statistical analysis and artificial neural networks (ANN); resulting in that according to the correlation coefficient; for exit humidity the RNA obtained R = 0.985; with a Mean Square Error (MSE) of 0.13; while in the statistical analysis an R = 0.853 was obtained with a mean square error of 0.5. On the other hand, for undulation the RNA obtained R = 0.955; with an MSE of 4.02; the statistical analysis obtained R = 0.753 with a standard deviation of 2.018; in this way, it is concluded that the prediction by neural network allows to obtain more exact and reliable results. Based on this analysis, future research is recommended to carry out a new statistical study with the elimination of the factors that reflect the highest P-values.