Non-contact electrical resistivity measuring system based on BP neural network

In order to reduce the complexity and compensate poor reliability of the resistivity measurement instrument for high temperature liquid metal, a set of non-contact electrical resistivity measuring system is designed based on electromagnetic induction principle and BP neural network. It consists of four components: the eddy current probe coil, the impedance analyzer, the temperature acquisition device and the related software. A finite element simulation model of probe coil is established in COMSOL. BP neural network model is implemented to approximate the implicit mapping between the reactance variation of coil and the resistivity of the metal, since the relationship is difficult to be explicitly expressed by the combination of elementary functions. Depending on BP neural network, resistivity can be predicted from the data of coil collected by the system. The qualitative measurement of liquid zinc's resistivity during cooling process is realized. The results clearly reflect the properties of resistivity in liquid state as well as solid state. The resistivity mutation phenomenon during phase change period is observed. The experimental results indicate that the system can be applied to resistivity measurement of metal in liquid and solid state with considerable reliability and detection capability.