Modeling Plasma Modification of Glass Surface using Neural Networks

In order to understand the relationship between discharge plasma parameters and material surface properties, neural networks model was constructed. The sample data were yielded from a lot of experiments for glass surface hydrophobic treatment. The experiments were arranged according to uniform design method and conducted using atmospheric pressure dielectric barrier discharge (DBD). Three discharge parameters, such as voltage, frequency and treatment time, were viewed as the inputs of the model. The outputs of the model were glass surface water contact angles. Back-propagation algorithm was used to train neural networks model. Model evaluation was carried out by simulation and error analysis. The optimized model was applied to predict, and the results are in agreement with practical situation. The obtained neural networks model has excellent predictive capability