Bolt axial stress measurement method based on ultrasound multi-feature fusion

There exist issues of ill-adaptability and low-accuracy in ultrasonic bolt stress measurement based on acoustic elasticity and scattering attenuation. Therefore, a bolt axial stress measurement method based on ultrasound multi-feature fusion is proposed in this paper. Firstly, ultrasound characteristic parameters based on the theory of acoustic elasticity are extracted, including the difference of time-of-flight (TOF) and the longitudinal wave attenuation coefficient in polycrystalline materials within the Rayleigh scattering range. Then, following the dimension reduction principle, the difference of TOF, attenuation coefficient, and clamping length are chosen as the input feature vectors for the fusion model. Subsequently, the SVR model is employed to establish a multi-feature fusion model for bolt axial stress measurement. Calibration and measurement experiments are conducted using the constructed experimental platform. Comparative analysis between the traditional measurement methods and the proposed method reveals that the SVR model can adapt to measurements of bolts with different lengths, overcoming the limitations of traditional methods. Furthermore, the measurement results of different fusion algorithms show that SVR yields significantly lower average relative errors in the measurement of the three bolts (3.35%, 4.44%, and 4.48%) compared to BP and RBF algorithms. This indicates higher measurement accuracy and immunity to low-stress effects.