Modeling approach for distributed RF MEMS phase shifter based on IA-BP neural network

An Efficient modeling technique based on immune algorithm and BP neural network is presented for the design of RF MEMS phase shifter. Three sensitive parameters are selected according to complicated three-dimensional structure design of an RF MEMS phase shifter and used as inputs of neural network. In the model, immune algorithm is first used for global search and then BP algorithm for local search. Experiments show that the proposed approach in this paper is a high efficiency modeling for the RF characteristics analysis for up/down-state of RF MEMS phase shifter. Comparison between improved BP neural network predictions and HFSS simulations show that the root mean square errors, mean absolute errors and maximize absolute errors are less than 1.77dB(5.34o), 2.24dB(5.93o) and 2.60dB(6.19o) respectively. Also, improved BP neural network reduces the training time at least 30 minutes.