Parameter estimation of multi-wavelength interdigital sensors based on optimized neural network

As a type of novel capacitive sensor, the interdigital sensors are widely used in non-destructive measurement of material properties in industrial process control. The multi-wavelength interdigital sensors have multiple penetration depths and can be used for measurement of multilayer material properties at different depths from the surface. Because of the nonlinear characteristic of interdigital sensors, the inverse problem of estimating material properties is complicated. In this paper, artificial neural networks are used for parameter estimation. The genetic algorithm optimized back propagation neural networks are proposed. First, the output of sensor is simulated with finite-element software over the entire range of input parameters. Then, the neural networks are trained to estimate the permittivities. The three-wavelength sensor is fabricated for the multilayer sample measurement. The results show that the genetic algorithm optimized back propagation neural networks algorithm can implement stratified analysis effectively.