Evaluating the optimization of irradiation components of mung bean seeds with ultrasound for increased seedling vigor, using artificial neural network

Extended Abstract   Introduction: A large number of experimental evidence indicates the positive effect of irradiating the seed with ultrasonic waves; so that irradiation causes the production of a more vigorous seedling. Conversely, inappropriate intensity and duration of irradiation can impose deleterious effects on seedlings by damaging the enzymatic activity. There are complex inter-and intra-relations between irradiation components (pre-soaking duration, temperature, and duration of irradiation) and response variables [seedling dry weight (SDW) and percent of abnormal seedlings (PAS)]. Therefore the balance values of the irradiation components cannot be precisely obtained by mean comparison. This study aimed to optimize (finding the balance values of) irradiation components for increased SDW, but diminished PAS of mung bean, using an artificial neural network.  Materials and Methods: A factorial experiment was conducted based on a completely randomized design with three replications. The factors were six pre-soaking durations (2, 4, 6, 8, 10, and 12 hours), 5 irradiation durations (0, 3, 6, 9 and 12 minutes), and 4 irradiation temperatures (17, 22, 27, and 32 oC). The 25 seeds were chosen for each petri dish. The multi-layer perceptron neural network was used to quantify the relations between variables; the experimental factors were used as the input (regressors), and PAS and SDW as the output of the model (response variables). Results: The analysis of variance results indicated that the simple and interactive effects of factors were significant on PAS and SDW. The structure 3:3:2 of the neural network, which is based on Secant Hyperbolic function, was suitable. The SDW and PAS were negligibly different for the contribution of the factors in determining their changes. In terms of relative contribution, the factors ranked from higher to lower as irradiation duration, irradiation temperature, and pre-soaking duration. The optimized values of components of irradiation by the neural network were irradiation temperature of 17.96 oC, irradiation duration of 5.3 minutes, and pre-soaking duration of 11.25 hours. For these components, SDW was 27% higher, and PAS tended to be 0.6% lower, compared to the best component combination gotten by mean comparison. Conclusion: Due to the highly strong interaction of irradiation components on seedling growth, the effect of component (s) tends to be changed intensively with changing the quantity of each component. In terms of finding the best combination of irradiation components, the neural network was more efficient than the mean comparison. Therefore, the neural network could be used as a complementary procedure in such investigations. Highlights: 1- Irradiation components including irradiation duration and temperature, and pre-soaking duration affected seedling growth. 2- Inappropriate irradiation components diminished seedling growth to the below of no-irradiation conditions. 3- The optimum (balanced) levels of irradiation components increased seedling growth remarkably.