Study On Design And Inverse Problem Of Planar Electromagnetic Sensor

Metal parts of industrial equipment are prone to produce cracks,corrosion and other damage because they are exposed to harsh environment for a long period of time.The equipment safe operation can be ensured by detecting the damage timely and accurately.The metal material damage detection method based on planar electromagnetic sensor is studied in this thesis.Three-dimension model of planar electromagnetic sensor is constructed based on Ansys Maxwell.The excitation coil structure is determined by analyzing the magnetic field distribution.The relationships among structure parameters and signal intensity,sensitivity are analyzed.The structure parameter optimization method based on response surface and genetic algorithm is proposed.The mapping relationships among output impedance of the sensor and the conductivity and liftoff of the material are analyzed.The inverse problem model is set up based on Least-squares polynomial fitting algorithm and BP neural network algorithm with particle swarm optimization.The result shows that BP neural network algorithm with particle swarm optimization has better prediction accuracy.The relationship between minimum phase and liftoff is analyzed by frequency scanning.The reciprocal function model for liftoff and minimum phase is given.The flexible planar electromagnetic sensor is fabricated.The calibration show that the measured values of planar electromagnetic sensor are in good agreement with the simulation,the calibration errors are small and the sensor has a good repeatability.The conductivity measurement experiments are implemented with different materials of 1060 pure aluminum,6061 aluminum alloy,2A12-T4 aluminum alloy and 7075 aluminum alloy.The results show that when the conductivity of the material is less than 25MS/m,the measurement system based on neural network algorithm has high measurement accuracy.The scanning imaging of planar electromagnetic sensor is studied base on the constant step and the sensor unit step separately.The damage scanning images are plotted based on the impedance amplitude.The damage images based on conductivity value are obtained by using neural network.The iterative threshold and the maximum variance between classes are used to separate the damage scanning images.The method effectiveness is validated by simulation.