To solve the problems of low resolution, weak anti-noise performance, and imprecise fault location in discharge recognition of ultraviolet imaging, an intelligent state recognition method of surface discharge based on visible images and machine learning is proposed in this paper. A visible image library of surface discharge(AC) under different experimental conditions was established, and the chromatic, grayscale, and morphological features of images were extracted. The image library was divided into 4 stages by clustering. Combined with spectrum correlation experiments, the physical meaning of the division results was explained. Supervised learning algorithms, including classic algorithms and deep learning, were used to train intelligent recognition models. The results showed that models based on chromatic features had a higher accuracy for recognition; the accuracy of artificial neural networks reached 0.982, which was obviously higher than that of classical learning algorithms (0.886). The results show that visible images can be used for the state recognition of surface discharges effectively. The application of machine learning ensures high efficiency and accuracy. This method will enable state recognition and fault location to be completed simultaneously.

As the uniform dielectric barrier discharges (DBDs) consisted by a large number of microdischarges in a long timescale may meet the needs of industrial application, the evaluation of the uniformity of DBDs used in the industry is different from that in the laboratory. In this paper, a digital image processing method presented in our previous paper is used to describe the uniformity of DBDs caused by the rotary plate electrode in a long timescale. Whether or not the electrode is rotated, with an increase in time, the gray-level standard deviation Std of the discharge image gradually decreases and reaches its stable value at last. The electrode rotation results in reducing the time constant required to reach a steady state for the gray-level distribution. With an increase in rotary speed, Std gradually decreases and reaches its stable values at last. This means that the electrode rotation improves the uniformity of the discharge. The reason produced the aforementioned phenomena should come from the relative motion of the accumulated charge on the dielectric surface and microdischarge remnants in the volume.

Can we detect electric discharge states in gases based on the information on visual images? This article proposes a new kind of method where we build several detection models for different states of corona discharge by applying four kinds of machine learning algorithms to extract color, brightness, and shape information characteristics of visible images taken by a digital camera. Every model is then tested on a new set of images to measure its performance. The four different machine learning algorithms are support vector machine (SVM), K-nearest neighbor regression (KNN), single layer perceptron (SLP), and decision tree (DT) algorithms. The prediction results show that the color features perform best among all three types of features and the KNN algorithm performs best among all four algorithms. This article also presents a discussion on how to choose the optimal detection areas of images for better detection performance. Our approach shows consistent results across different cameras and camera settings. The results demonstrate that even if only the visible light spectrum emitted from a plasma is captured, the color method can provide sufficient discharge information for economic and convenient use in discharge state detection because the species producing visible radiation are affected by radiation in all bands.

Discharge is a process, which involves changes in multiple species, and it is important for researchers to be able to describe changes in discharge status holistically. In this study, we applied AC and DC high voltage power sources to a short-gap needle-plane device in an atmospheric environment. We propose two new corona diagnosis methods based on spectral data and visible images. The experimental results demonstrate that, in the RGB system, the proportion of the blue component in AC corona images has the characteristic of decrease-stable-increase with voltage, and the corresponding result of the spectra is stable-increase. In the XYZ system, the AC corona chromaticity coordinates of both images and spectra demonstrate a linear fitting result with the voltage, and the image results contain a retracement process, showing shows the increase-stable-decrease trend of chromaticity coordinates x and y, which correlates well with oscilloscope results. The difference between the image and spectra results indicates that, compared to the spectra, high-resolution visible images are more sensitive to imperceptible changes in colorimetry information. The change of DC corona colorimetry characteristics can illustrate the change in AC corona colorimetry characteristics. Additionally, the dominant wavelength is proven to reflect the discharge species, and can reflect the vibrational temperatures. To the best of our knowledge, this paper represents the first attempt to systematically introduce colorimetry theory into discharge images and spectra. We successfully obtained colorimetric indicators with the linearity and low-dimension characteristics, which correspond well with known physical parameters.

The noncontact diagnosis methods of discharge plasma include spectrum, electromagnetic wave, ultraviolet image, and high-speed nanosecond image analyses, but there are few studies on the visible image diagnosis method. With the development of digital cameras, the number of pixels in a visible image is far greater than that of ultraviolet, infrared, and high-speed cameras; the image contains a considerable amount of discharge spatial information and provides a new solution for plasma diagnosis. We obtained visible digital images of short gap negative dc coronas with the corresponding oscilloscope current signals under the experimental conditions of atmospheric pressure and room temperature. Two colorimetric indicators (RGB and HSI) were proposed to study the colorimetry features of discharge images, of which the analysis was correlated with the amplitude and frequency of oscilloscope current signals. The result shows that the hue value of the overall image region is positively related to the pulse amplitude, while the intensity value corresponds to the pulse frequency. The stage division based on the hue value is better than that based on current pulses, which shows that the Trichel stage can be further subdivided. In addition, we observed that the anode discharge begins earlier than the formation of pulseless glow, and we illustrated that anode discharge played an important role in the formation of pulseless glow and spark breakdown. Briefly, the discharge image corresponds well with the results of traditional detection methods and provides a new perspective for the corona discharge study, deserving further research.