It is usually to see a large number of pipelines separating around the refineries, chemical and petro-chemical plants. The corrosion and erosion defects are unavoidable to occur in transporting pipe line. Especially, the maintain stuff usually find out breakage pipe or leaking liquid at elbowing pipe line because of the corrosion and erosion defects. So it is essential to examine these pipelines with an efficient method. The use of guided waves method is very attractive to solve this problem since guided wave could be excited at one circle on the pipeline and propagate over considerable distance. To choose guided wave torsion mode T (0, 1) as excitation mode because its group velocity doesn¡¦t change with frequencies. And the research analyzes the mode conversion that occurred when T (0, 1) mode propagated after the elbow pipe. The research also discusses the signal difference in different depth, circumferential distribution and axial length defects on the elbow pipe. The erosion defect usually occurs in the elbow pipe line and it would change with fluid velocity, causticity of fluid and flow direction. Therefore, the research designs the defects according to the character of erosion defect by finite element method software and simulates T (0, 1) mode propagating in the pipe line. Then this research extracts and analyzes the reflection signals from defects. In this guided wave experiment, the research manufactures the defect on elbow pipe. Because the erosion defect could be usually found at outer side of elbow pipe, artificial defect would be set there. And the elbow pipe is manufactured with different depth, circumferential distribution and axial length defect. The research would discuss the relationship between change of defect and reflection signal. By elbow pipe defect signals of simulation and experiment consequence, the different depth, circumferential distribution and axial length defect signals could be still distinguished. The signals with different axial length defect that received from straight pipe and elbow pipe are similar and are affected by signal constructive and destructive interference. So the research could get maximum and minimum defect signal amplitudes from one-fourth wavelength axial defect and half wavelength axial defect. Therefore, the axial length defect of elbow pipe could be estimated from defect signals and this consequence could help judge the level of damaged elbow pipe. T (0, 1) mode has better sensitivity to outside of the pipe than inside of the pipe. So the bigger signal amplitude could be received from the notch at outside of the pipe. In the process of wave propagation simulation, there are overlapping waveforms and mode conversions occur at elbow pipe. This situation causes the defect signals were amplified at elbow pipe. In practical detection, the misjudgments of amplified defect signals should be attended to.