研究目的
To classify fault types in electrofusion polyethylene joints using machine learning, thermal pulsing, and IR thermography methods to enhance the security of polyethylene gas pipelines.
研究成果
GLMNet outperformed k-means and Random Forests with a classification accuracy of 93.75%. Supervised methods showed better performance than unsupervised methods. Increasing the data bank size improved classification accuracy. The study highlights the importance of machine learning techniques in fault classification for enhancing pipeline security.
研究不足
The study is limited by the sample size and the types of faults considered (ovality and misalignment). The accuracy of classification methods varies with the number of samples and the type of fault.
1：Experimental Design and Method Selection:
The study extends previous experimental IR-thermography data bank and applies k-means, Random Forests, and GLMNet algorithms in a two-stage approach for fault classification.
2：Sample Selection and Data Sources:
Joints were deliberately prepared with ovality and misalignment faults. Thermal images were captured during the electrofusion process.
3：List of Experimental Equipment and Materials:
Electrofusion machine, IR thermal imaging camera, PE pipes, and couplers.
4：Experimental Procedures and Operational Workflow:
Electrical pulses were induced for 5 seconds, converted to heat pulses, and the temperature profile was recorded. Images were pre-processed for fault classification.
5：Data Analysis Methods:
k-means, Random Forests, and GLMNet algorithms were used for classification accuracy assessment.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
