Monitoring of Arc Plasma Process Parameter Using CNN-Based Deep Learning Algorithm to Accommodate Sensor Failure

Abstract

Redundant soft sensors are used to provide information on physical parameters in industrial manufacturing processes to accommodate conventional sensor failure. In this article, a convolutional neural network (CNN)-based deep learning method is proposed based on image processing to estimate the process condition of a transferred arc plasma from visual images. The proposed method adds redundancy in sensing a high-temperature smelting process so that both arc current and gas flow rate can be estimated indirectly from the images of the plasma glow. The visual images of the different experimental processes were trained in a new customized CNN model and the classification performance of the proposed model is also compared with five well-known CNN-based deep learning architectures, such as AlexNet, SqueezeNet, InceptionV3, DenseNet121, and ResNet101V2. The classification of process parameters through images from a deep learning model can be used for the immediate detection of any change in source current and gas flow rate when there is a failure of the gas sensor or current sensor.