A Method for Hydrophobicity Detection of Composite Insulators Based on Lightweight Convolutional Neural Network

Abstract

The invention discloses a method for detecting the hydrophobicity of a composite insulator based on a lightweight convolutional neural network. Firstly, a composite insulator water spray image sample set containing seven levels of hydrophobicity is constructed and labeled, and the sample set is expanded through preprocessing and collected. Divide the training, verification and test sets; then migrate the lightweight convolutional neural network models trained in large data sets such as MobileNet, ShuffleNet and GhostNet, modify the model structure and output, set the learning rate in segments for the network layer, and construct Hydrophobicity detection model, and optimize the model through Adam, SGDM and other algorithms; finally, use the composite insulator hydrophobicity grade intelligent recognition model to detect the hydrophobicity of the water spray images in the test set, and output the recognition results and accuracy. The invention can overcome the limitation that the traditional detection method needs to manually crop pictures and is seriously affected by light, and can improve the efficiency and accuracy of the hydrophobicity detection of the composite insulator.

Description

technical field [0001] The invention relates to the technical field of hydrophobicity detection of composite insulators, in particular to a method for detecting the hydrophobicity of composite insulators based on a lightweight convolutional neural network. Background technique [0002] Composite insulators have good pollution flashover resistance due to their surface hydrophobicity and hydrophobic migration. However, during long-term operation, the silicone rubber material will age, resulting in a decrease in hydrophobicity. In engineering, the water spray classification method (HC method) is usually used to carry out random inspection of the hydrophobicity level of the running composite insulator, and determine the inspection cycle and determine whether it can continue to operate according to the test results. However, the judgment result of HC level depends on the subjective understanding of the tester on the water spray image, which is easy to cause misjudgment. The hydr...

AI Technical Summary

Problems solved by technology

The accuracy and high robustness of CNN to detect the hydrophobicity of composite insulators depends on the depth of the network, which requires high storage and calculation of hardware devices, and is difficult to apply to mobile devices with limited storage and calculation.

Images