中文摘要：

采用探地雷达技术、超声波透射法等多项技术方法对T梁的压浆质量进行试验性检测研究.首先采用超声波透射法在未压浆的钢束管道位置和正常混凝土位置（即避开波纹管位置）进行测试,然后将这两种类型的测试数据结果作为学习样本提供给神经网络进行学习;再将已压浆预应力T梁上的实测结果作为网络测试数据进行计算,根据神经网络的计算结果可预测钢束管道内的压浆质量,判断管道中是否有空隙存在.根据检测结果和开窗验证的情况来看,超声波透射法检测和神经网络的预测结果是准确可靠的,探地雷达技术不适合于T梁束孔管道的压浆质量检测.

英文摘要：

Pipelines of prestressed concrete beams should be grouted fully so as to keep beam entity. There were few researchers to study how to test the grouting quality of concrete beams all over the world. The authors had done the experimental study and detected beams＇ grouting quality by ground penetrating radar, ultrasonic pulse-veloclty method, etc. The characters of ultrasonic wave in an un-grouted beam could be tested by ultrasonic pulse-velocity method at first. The testing positions should go through corrugated shells vertically or avoid them. There are crannies when testing positions go through a corrugated shells vertically because of un-grouted cause, otherwise, there are not crannies. Then, taking the testing data as training elements of a feed-forward back-propagation network, and the grouting quality of a grouted beam can be predicted or judged by a trained network. Some test positions have been validated after the experimental test. It indicates that the predicting results of a feed-forward back-propagation network are correct, but the technology of ground penetrating radar is unfeasible for the grouting quality testing.