Research on Fault Diagnosis Technology for Ion Tube Based on Neural Network

Abstract

Ion tube is a type of dielectric barrier discharge (DBD) device widely applied in the field of air pollution treatment. The operational state of ion tube is a crucial factor that affects the efficiency of air pollution treatment. Manual inspection is the primary method for monitoring the state of ion tube. This approach suffers from issues such as time-consuming, labor-intensive, and heavily reliant on personal experience. To achieve automation and intelligence of fault diagnosis for ion tube, this article uses a neural network to design an online measurement scheme of ion tube’s Lissajous figure, the nonlinear relationship of signals from the low-voltage side to high-voltage side of the ion tube transformer is fit by neural network. And based on the measured low-voltage side signals, the ion tube’s Lissajous figure is calculated by the designed neural network. Moreover, the convolutional neural network (CNN) is used to construct the fault diagnosis scheme for ion tube and the ion tube’s Lissajous figure is classified by a two-level classification scheme. The primary classification CNN distinguishes between punctured and nonpunctured ion tube, and then the secondary classification CNN categorizes nonpunctured ion tube into brand-new, semi-new, and damaged ion tube. The experimental results indicate that the designed online measurement technology of ion tube’s Lissajous figure has the same measurement accuracy as traditional methods and does not require oscilloscopes, high-voltage probes, and external measurement capacitors. And the designed fault diagnosis technology for ion tube effectively distinguishes four fault states of ion tube with high accuracy.