Research On Defect Recognition Method Of Pipe With Cladding Based On Improved ACO-SVM

The continuous construction of oil and gas pipelines in China has put forward higher requirements for its reliability and safety,of which pipelines with cladding account for a considerable proportion.In order to ensure the safe operation of the pipeline and avoid major safety accidents,it is necessary to conduct regular defect detection and analysis on the pipeline.In this thesis,based on the pulse magnetic eddy current detection method,the defects of pipelines with cladding layers are detected,and the detection signals are de-noised,feature extracted and classified.Intelligent optimization algorithms are used to optimize the parameters of the support vector machine model,and the ant colony algorithm is improved.The experiment proves that the classification accuracy and stability of the support vector machine model optimized by the improved ant colony algorithm have been improved.The main contents of this thesis are as follows:(1)According to the characteristics of the detection signals of pipeline defects with cladding,the basic wavelet functions of db3,db5 and sym5 were used to denoise the signal.The results show that the basic wavelet function of sym5 is the best for denoising the square wave pulse signal with a signal-to-noise ratio of 25.0186 d B.(2)Compare the classification effect of SVM model optimized by different algorithms on wine data set.Select RBF kernel function,use ant colony algorithm,genetic algorithm and particle swarm algorithm respectively to find the best SVM parameters,ant colony algorithm optimized SVM model achieved the best classification results with94.96% accuracy.(3)Propose an improved ant colony algorithm.Based on the traditional ant colony algorithm,the algorithm dynamically adjusts the pheromone increment,uses a variable step size grid search ant colony algorithm related to the number of iterations,and introduces the genetic algorithm mutation operation to reduce the probability of ant colony algorithm falling into local optimum.After the detection signal is denoised by wavelet threshold,38-dimensional feature data is extracted and the principal component analysis method is used to reduce the dimension.Using the improved ant colony algorithm to optimize the SVM model to classify the feature data,compared with the algorithm before the improvement,the running time of the improved model increased by11.53%,the accuracy increased from 78.41% to 80.49%,and the stability of the algorithm was improved by 30%.(4)This thesis establishes a pipeline defect signal database system based on SQL Server and C #.The database contains information about pipeline inspection data,it has the functions of querying,adding,modifying,and deleting related information such as the detection date,location,and defect type of the detection signal,which is conducive to the digital construction of pipeline management.