基于特征量和神经网络的钢管缺陷预测模型

分析了钢管缺陷几何大小与缺陷漏磁信号(MFL)特征量之间关系，建立了一组全方位的钢管缺陷信号特征量，并将人工神经网络理论和算法应用于钢管缺陷预测。通过实验取得样本，在对网络进行训练的基础上，建立了基于钢管缺陷漏磁信号特征量和神经网络的缺陷预测模型，继而根据漏磁信号对缺陷进行定量预测。给出了实验结果，结果表明采用这种方法能够较好地实现管道缺陷的定量识别。

Prediction Model for Steel Pipe Defects Based on ANN and Characteristic Values

The paper analyzes the relationship between the magnetic flux leakage(MFL) signals and the dimension of steel pipe defects A series of the characteristic values of MFL are defined and the algorithm of arti f icial neural network(ANN) are applied in the research for estimation to The dimension of pipe defects The samples are acquired by experiments On the bas is of the training of the network, the prediction model based on the characteristi c values of MFL and ANN is built up The dimension of pipe defects can be quant itative estimated by the model The experiment results show that the new method is quite effective for quantitative recognitio n of steel pipe defects