基于声发射技术的箱型梁大应力区故障模式识别

在介绍了声发射技术的基础上，采用BP神经网络，分别对预置故障的港口起重机典型结构——箱型梁的声发射信号进行分析。选用原始波形特征参数以及它们之间组合派生出的特征参数等9个参数作为神经网络的输入，设计箱型梁大应力区常见五种故障模式：正常、局部表面裂纹、局部深埋裂纹、局部焊缝、对接焊缝为最终识别分类模式。结果表明，该方法可以对以上五种模式的声发射源进行有效识别。

Faults Pattern Identification of Box Girder Major Stress Zone Based on Acoustic Emission Technique

On the basis of introduction of acoustic emission technique, and by adopting BP neural network, the a- coustic emission signals acquired from the typical structure of port crane box girder which had been preset faults have been analyzed. Selection of the nine parameters of original acoustic emission waveform characteristic parameters and the characteristic parameters derived from their combination were taken as input to neural network. Common five failure modes for design of box girder major stress areas are normal, local surface cracks, local buried cracks, local weld, butt weld, which are identified the final classification mode. The resuhs showed that the method can effectively identify the acoustic emission source of the above five modes.