Modeling Conidiospore Production of Trichoderma harzianum Using Artificial Neural Networks and Response Surface Methodology

Abstract

An alternative to facing plagues without affecting ecosystems is the use of biocontrols that keep crops free of harmful organisms. There are some studies showing the use of conidiospores of Trichoderma harzianum as a medium for the biological control of plagues. To find the optimal parameters to maximize the production of conidiospores of Trichoderma harzianum in barley straw, this process is modeled in this work through artificial neural networks and response surface modeling. The data used in this modeling include the amount of conidiospores in grams per milliliter, the culture time from 48 to 136 h in intervals of 8 h, and humidity percentages of 70%, 75%, and 80%. The surface response model presents R2 = 0.8284 and an RMSE of 4.6481. On the other hand, the artificial neural network with the best performance shows R2 = 0.9952 and RMSE = 0.7725. The modeling through both methodologies can represent the behavior of the Trichoderma harzianum conidiospores growth in barley straw, showing that the artificial neural network has better goodness of fit than the response surface methodology, and it can be used for obtaining the optimal values for producing conidiospores.