Defective Multilayer Composite Pipes Containing Large Thickness Polyurethane Ultrasonic Detection

With the rapid urbanization and improvement of infrastructure construction in China,buildings are becoming more centralized and large-scale.The district heating(DH)industry has become one of the first choices for energy supply for large-scale building clusters in northern cities and towns in China due to its relatively energy-saving and environment-friendly advantages.In the past decade,the length of China’s heating network has more than tripled,and hundreds of thousands of kilometers of district heating pipes(DHP)face the threat of continuous aging,especially the insulation layer of the pipes may stick off,fracture and other serious accidents once they are over-aged.An important means to effectively prevent premature over-aging of pipelines is to detect cavitation defects in the insulation layer of pipes that have not yet left the factory,so it is necessary to study a set of methods that can quantitatively detect internal defects in DHP,which has great theoretical and practical significance for energy conservation and environmental protection,urbanization construction,and improving the quality of life of residents.The goal of this paper is to investigate a set of ultrasonic inspection equipment that can be carried by inspectors for internal defect detection of DHP.Ultrasonic inspection technology(UDT)is the inspection tool used,and cavitation defects in the polyurethane(PUR)layer of the DHP are the object of inspection.Due to the large thickness of the PUR layer and the complex interface information of the multilayer structure of the DHP,quantitative detection of porosity defects is a difficult task.To solve this problem,a detection method combining feature extraction,feature selection and crow search algorithm optimized support vector machine(CSA-SVM)is proposed in this paper.Firstly,by studying the acoustic properties of DHP,the main parameters of ultrasonic detection and detection method are designed to obtain the original sample set by using ultrasonic transmission method and oblique incident reflection method for defect signal acquisition of prefabricated cavitation defects of different size-ranges on the basis of controlled environment,parameter variables.Secondly,the original signal is analyzed in time and frequency,and a filter feature selection method for de-redundancy and relevance enhancement is proposed based on the specificity of the feature set: 1)Based on Chatterjee’s rank correlation coefficient and CRITIC weights,the approximate Markov blanket theory is used to provide the de-redundancy decision.2)By the Relief-F algorithm,the relevance ranking of the de-redundant features.3)A strategy based on the feature selection method to obtain the subset with the minimum redundancy and maximum relevance from the original feature set;4)multiple methods are used to verify the enhancement brought by the feature selection method to the detection process.The results show that the proposed feature selection method is not only positive for feature interpretability of defects and cavitation defect recognition accuracy,but also can be used to reduce hardware load and training time.Finally,three commonly used classifiers are used to compare with the proposed CSA-SVM.The experimental results show that CSA-SVM has the highest accuracy in defect size prediction,and experiments using out-of-sample data validate the generalization and feasibility of the CSA-SVM classifier.All experiments show that our proposed set of methods can well solve the problem of quantitative detection of cavitation defects in insulation layer of DHP,which is expected to provide a sustainable and healthy development in the DH industry.