Ultrasonic Echo Parameter Estimation And Its Application For Flaw Detection In Pipeline

Pipeline is the main carrier of energy transportation. Due to the influence of environmental stress, such as corrosion, temperature and pressure, it’s easy to produce defects and lead to leakage and other accidents which would result in an immeasurable loss. Therefore, regular inspection of the pipeline has great significance. The key problem of quantitative estimation of defects is how to estimate accurate parameters of flaw from inspection data. The main aim of this thesis is researching the parameter estimation methods based on ultrasonic echo signal model from the traditional sampling data and sparse sampling data, and applying them to pipeline ultrasonic testing.Traditional sampling using the Nyquist sampling theory is the primary method on ultrasonic signal at present, which possesses the features of massive sampling data and integrated signal information. However, it is not suitable in the cases of multi-transducer array, long time and long distance testing, because of the whole data storage and analyzing bottleneck. Recently presented sub-Nyquist sampling method can solve the problem, but cannot directly estimate the parameters from the sparse sampling data. Considering the concurrent status of the two kinds of sampling methods, we carry on research of parameter estimation methods from two kinds of sampling data to provide references for pipeline flaw ultrasonic testing and evaluation.To the regular sampling data, the signal parameters are estimated by using parameterized model based method. Meanwhile, a hybrid parameter estimation method has been proposed and researched based on LM(Levenberg-Marquardt) algorithm and LP(Linear Prediction) algorithm for the purposes of solving the problems of heavy computation and prior original value. To the sparse sampling data based on FRI(Finite Rate of Innovation) theory, parameter estimation method suitable for more parameters estimation is proposed. Meanwhile, a DTWM(Dynamic Time Window Modulation) method is developed for improving the estimation accuracy on noisy condition.The main research works of this thesis are as follows:(1) Based on analyzing the parameter estimation methods from traditional sampling data, a hybrid algorith combined with LM and LP is proposed aiming at solving the problems of original value selection and heavy computation.(2) An ultrasonic envelope model based on generalized Gamma distribution is established, which can fit asymmetric echo envelope. Comparing with asymmetric Gauss envelope model, this model can reduce the amount of the estimation parameters. The performance of this model has been verified through experiments.(3) Based on analyzing the state-of-the-art of the parameter estimation method using FRI sparse sampling data, four kinds of parameters, which are arrival time, peak amplitude, pulse width and the number of pulse train, have been accurately estimated using single channel FRI sampling data obtained from pipeline testing.(4) Considering the problems of convenient hardware implementation and estimation precision in low SNR, a further research on parameter estimation method has been done by using multi-channel FRI sparse sampling data. DTWM method has been proposed and researched to improve the estimation accuracy with the performance verified by real experiments.