A Generalized Approach for Failure Detection on Steel Products with the Use of an Eddy-Current Sensor

Abstract

In this paper we present a signal processing method used for the improvement of the performance of an eddy-current sensor. This method is applied on different processes of steel industry-(steel slab production, tube soldering) in order to detect surface or welding defectsThe proposed method is based on the modelling of the signal without failure (M0 model) and with failure (M1 model). As the signal is time-varying, a simple failure detection approach only based on the knowledge of MO or Ml model is not sufficient. For efficiency it is necessary to have an Ml adaptive model, which requires the use of MO model. The proposed strategy uses a combination of these two modelsFailure detection is discovered with a failure-sensitive Kalman filter (M0 model) and a Hinkley detector coupled with a basic recognition algorithm which gives the duration of failure. This parameter is used to adapt M1 model and then a standard deviation test of failure magnitude estimation, obtained with a second Kalman filter (Ml model) permits the confirmation of the failure