| Summary: | This paper describes a new method of eddy current impedance extraction and hardness classification for valves, which is used in online classification. A finite element simulation is conducted to estimate the base magnetic response in specimens with various initial permeabilities, and a scatter diagram between the hardness calibration and initial permeability is plotted experimentally; thus, the hardness classification model is obtained by using the initial permeability as the intermediate variable. A lock-in amplifier is commonly used to measure the amplitudes of impedance components in conventional eddy current testing. This study replaces the lock-in amplifier with a cross-correlation algorithm with the objective of integrating the impedance processing with the subsequent classification. Moreover, an incremental and decremental support vector machine is used to classify the valve specimens based on the extracted impedance due to its demonstrated excellent performance for both linearly separable and nonlinearly indivisible problems. The cross-correlation algorithm and the incremental and decremental support vector machine are implemented on a field-programmable gate array, which has the advantage of parallel processing and can satisfy the online requirements. A comparison with the traditional impedance comparison method shows that the proposed method not only is suitable for online hardness classification but also achieves good performance in correlating impedance amplitude changes caused by a temperature increase of the coil and various heat treatment conditions with the hardness values.
|
|---|