论文部分内容阅读
目前钢轨平直度测量仪检测精度评定方法存在大量测量仪的检测数据未参与其检测精度评定的数据统计分析中,为了能精确评定测量仪检测精度,需更多千分尺测量数据而增加人工测量的工作量。采用BP神经网络基于千分尺测量得到仿形面曲线函数B0(n)建立仿形面波形数据数学模型,并基于BP模型残差分析引入周期性函数进行组合修正,以得到千分尺在各个对应的测量仪检测位置上的测量值的曲线函数B(n),通过对比分析测量仪的连续检测数据A(n)与B(n)的相似度,可更充分的对钢轨平直度测量仪的检测精度进行评定及其性能的全面验证。实验结果表明:组合模型建模方法正确可行、建模精度较高,其平均绝对误差仅为0.0218 mm,平均相对误差仅为1.09%。
Present rail flatness measuring instrument testing accuracy assessment method There is a large number of measuring instrument test data did not participate in the accuracy of its testing data statistical analysis, in order to accurately measure the measuring accuracy of the instrument, the need for more micrometer measurement data to increase the manual measurement Workload. BP neural network was used to establish the mathematical model of profiling surface waveform data B0 (n), which was obtained from the micrometer measurement. Based on the BP model residual analysis, a periodic function was introduced to modify the combination to obtain the micrometer in each corresponding measuring instrument The curve function B (n) of the measured value at the detection position can be more fully evaluated by comparing the similarity of the continuous detection data A (n) and B (n) of the measuring instrument to the detection accuracy of the rail flatness measuring instrument To assess and verify the performance of the full. The experimental results show that the combined model modeling method is correct and feasible and the modeling accuracy is high. The average absolute error is only 0.0218 mm and the average relative error is only 1.09%.