基于微磁检测和交替最小二乘法的钢轨裂纹定量估计

Quantitative Estimation of Rail Crack Based on Weak Magnetic Testing and Alternating Least Squares

中文关键词:  钢轨裂纹  弱磁检测  交替最小二乘  定量估计

英文关键词:rail cracks  weak magnetic testing  alternating least squares (ALS)  quantitative estimation

中文摘要:

      针对钢轨裂纹定量估计的问题，提出了一种基于弱磁检测和交替最小二乘法的钢轨裂纹定量估计方法。该方法使用弱磁检测的钢轨裂纹信号，提取弱磁信号特征，再利用交替最小二乘法解出钢轨裂纹的深度、宽度、水平角和垂直角的估计值。针对交替最小二乘分解大矩阵速度慢的问题，提出了优化初值的方法，减少交替最小二乘迭代次数，缩短计算时间。基于实测数据的研究结果表明，深度和垂直角估计精度较高，而宽度和水平角误差较大，从而提出将其他三个裂纹参数混入混合矩阵中，再去拟合深度、水平角和垂直角，估计精度将会明显提高。

英文摘要:

      In order to solve the problem of rail crack quantitative estimation,a quantitative estimation method of rail crack by weak magnetic detection and alternating least square (ALS) method is proposed.In this method,the rail crack signal detected by weak magnetic field was used to extract features,and the estimates of rail crack depth,width,horizontal angle and vertical angle were obtained by using ALS.In order to solve the problem of slow decomposition of large matrix by ALS,a method of optimizing initial value was proposed to reduce the iteration times of ALS and shorten the calculation time.The results of the measured data show that the estimation accuracy of depth and vertical angle is higher than that of width and horizontal angle.Therefore,it is proposed that the estimation accuracy of depth,horizontal angle and vertical angle will be significantly improved by incorporating the other three crack parameters into the mixed matrix and then fitting the depth,horizontal angle and vertical angle.