Identification of radiating trace configuration using finite element method and neural networks

The PCB traces produce electromagnetic field when excited by a signal source. The trace acts as a source of emission when impedance mismatch or a discontinuity along the length of the trace occurs which may interfere with other equipments placed in its closer proximity. In this paper a two stage feed forward neural network using Back Propagation algorithm is used to identify the radiating trace. The input parameters used for training the neural network is extracted from the thinned result of the spectrum obtained by Finite Element Analysis on the radiating structure. The thinned Pattern of type original radiating trace is compared with the trained pattern for correct identification of the radiating trace.