Application Of Intelligent Optimization Algorithm In Pipeline Leakage Detection

As an important means of transportation,pipeline plays a vital role in the development of modern society.With the increase of the scope and quantity of its application,pipeline will inevitably be damaged by various factors.Therefore,how to accurately judge whether the pipeline leakage occurs is particularly important in the case of noise.The Variational Mode Decomposition(VMD)algorithm is utilized in this paper for denoising of the pipeline signal,and the Support Vector Machine(SVM)is employed for classification and identification of signals from various working conditions of pipeline after denoising.Furthermore,the noise reduction capabilities of the VMD algorithm and the recognition accuracy of the SVM are improved through the adaptive selection of essential parameters using the enhanced Marine Predator Algorithm(MPA).The main work contents are as follows:Firstly,after careful study of the relevant knowledge of intelligent optimization algorithms,a comparison is made between the MPA,Genetic Algorithms(GA),Gravitational Search Algorithm(GSA),and Particle Swarm Optimization(PSO)using 19 benchmark test functions,in which mean value and standard deviation are used as evaluation metrics for the test functions.Based on the evaluation results,MPA shows faster convergence speed and stronger global search ability.Secondly,considering the commonly occurring problems of intelligent optimization algorithms including MPA itself,e.g.,the easily falling into local optima and the slow convergence speed,the Good point set strategy is used to increase the randomness of population initialization and improve the traversal of MPA search.The Dynamic Spiral Bubble method is used to enhance the predation ability of predators in MPA,which improves the convergence speed and search ability of local optimal value of MPA.The step length CF in MPA is improved to strengthen its ability to escape from local optimal values of MPA.The performance of the improved MPA(IMPA)is tested using the same benchmark test functions,and based on the evaluation results,IMPA has faster convergence speed and better global search ability compared to MPA.Thirdly,to address the problem that the decomposition accuracy of the VMD algorithm is seriously affected by the inappropriate selection of the decomposition level K and the penalty factor α in VMD algorithm,IMPA is used to optimize these two parameters and achieve adaptive selection of parameters.After performing IMPA-VMD on the signal,effective Intrinsic Mode Functions(IMFs)are separated by using Symbolic Dynamic Entropy(SDE)and then reconstructed.Subsequently,evaluation metrics such as Signal-to-Noise ratio,mean square error,and mean absolute error are used to analyze the performance.The results show that the IMPA-VMD method can significantly raise the decomposition accuracy of VMD.Finally,to address the issue of low classification recognition accuracy of Support Vector Machine(SVM),the IMPA-VMD method is used for denoise of the collected three types of pipeline signals,including normal,leak and knock,and reduction of the complexity of the pipeline signals.The kurtosis feature is extracted from the signal after noise reduction.As for the two important parameters of the radial basis kernel function in SVM,i.e.,the penalty factor C and the kernel function parameter g,which are difficult to artificially select,the IMPA-SVM method is used to find the optimal values for the two parameters.By comparing the lowest recognition accuracy,average recognition accuracy,and highest recognition accuracy,the effectiveness of the IMPA method is demonstrated in the aspect of parameter optimization of SVM.