结合人工神经网络和电磁仿真,给出了一种用于综合交指电容及Metal—Insulator—Metal(MIM)电容结构参数的方法。基于逆向神经网络,可有效地根据给定频点上的电容值快速准确地综合出其对应的结构参数,从而避免了反复优化的过程。同时,可以由训练好的神经网络参数得到结构参数相对于等效电容的闭式计算公式。数值结果验证了方法的正确性和有效性。
In this paper, a synthetic method based on Artificial Neural Network (ANN) and electromagnetic simulation is provided to directly calculate the structure parameters of interdigital capacitors and MIM capacitors with given the working frequency and corresponding capacitor's value. Since there is no searching process, the computing time can be greatly reduced once we got the trained reverse neural network. Besides, the close -form formulations can be easily obtained from the weight matrix, the bias matrix and the transfer functions. Numerical results show the efficiency and validity of this model.