Development Of Rail Stress Testing Instrument Based On Magnetic Barkhausen Noise Detection

In recent years, the domestic and international high-speed seamless railway has a rapiddevelopment, due to temperature stress is a major factor in the structure of health and seamlessrailway traffic safety, so how to detect railway temperature stress quickly and accurately becomesvery necessary and urgent. In this paper, based on the electromagnetism method,we build a railwaytemperature stress testing instrument to achieve accurate measurement of rail stress, convenient andfast.Based on the theory and engineering practice of Magnetic Barkhausen Noise detection ofdomestic and foreign,we set up a stress detection system, which includes the hardware platform,software platform and stress calibration and experimental validation. The hardware includes anexcitation signal generator module, a detection sensor module, a signal conditioning module, a signalacquisition module, a central processing module, a displacement sensor module and a consolecomputer control module, the software part is mainly composed of a data acquisition module, a dataprocessing module, a host computer module and a console computer control module.Before theinstrument calibration experiments, we determine the experimental excitation voltage and excitationfrequency,considering the resolution and sensitivity of the instrument,we use regression analysismethod for total data and segmentation data for the calibration of the instrument.Based on the sixeigenvalues of MBN signal changes with stress nonlinearly,we use a quadratic model in the totaldata,the difference of quadratic model and linear model is less in the segmentation data,so we use alinear model in the segmentation data.Because of the presence of human factors, experimental dataaffected by the sensor lift-off, it can be verified that automatic adjustment of excitation signalamplification and regression method can compensate for the sensor lift-off,we also explored theregression analysis method to compensate for the effect of temperature.The processor’s processingprecision is limited,therefore,to build the regression model we should first seek maximum linearindependence group of vector group,then remove outlier values in the process of regression, andfinally through the model validation,we find that error within a reasonable range.