基于BP神经网络的仿真线设计及其FPGA实现
doi: 10.3724/SP.J.1146.2005.01447
Simulation Line Design and Its FPGA Realization Based on BP Neural Network
-
摘要: 该文提出了一种采用BP神经网络实现仿真线的方法。首先采用遗传算法优化神经网络结构，用离线训练后的BP神经网络逼近传输线的传递函数，然后用STAM算法以较少的存储空间实现BP神经网络的激励函数近似，进而用FPGA和D/A转换器进行硬件实现。文中基于FPGA对长度为10000m，特性阻抗为55的同轴电缆进行了仿真线的硬件实现，实验结果验证了该方法的有效性。该方法可以推广到传递函数未知的传输网络的仿真应用中。Abstract: A new method for simulation line realization based on Back Propagation Neural Network (BP NN) is presented in the paper. Applying Genetic Algorithm (GA) to optimize the neural networks structure, BP NN is trained to correspond the transfer function of simulation line. Activation function of NN is approximated with STAM (Symmetric Table and Addition Method) algorithms. A coaxial-cable which is 10000m long and 55 line characteristic impedance is simulated and realized by using FPGA and D/A converter. Experimental results show that the proposed approach can greatly reduce the memory of hardware realization. This method can be generalized to simulate the transmission network with unknown transfer function.
-
中华人民共和国国家标准：常用电信设备名词术语GB1417-1978，北京: 技术标准出版社，1979.[2]Nielsen R H. Theory of the backpropagation neural network. Proceedings of the International Joint Conference on Neural Networks. Washington, USA. 1989, 1: 593-605.[3]李倩，王永县，朱友琴. 人工神经网络混合剪枝算法. 清华大学学报, 2005, 45(6): 831-834.[4]杨建国，翁善勇，赵虹等. 采用遗传算法优化的煤粉着火特性BP神经网络预测模型. 动力工程，2006, 26(1): 81-83.[5]陈作炳，艾春庭，夏雪峰. BP神经网络仿真软件. 计算机仿真, 2001, 18(4): 23-24.[6]Stine J E and Schulte M J. The symmetric table addition method for accurate function approximation[J].Journal of VLSI Signal Processing.1999, 21(2):167-177[7]Schulte M J and Stine J E. Accurate function approximations by symmetric table lookup and addition. Proceedings of the 11th International Conference on Application-Specific Systems, Architectures and Processors. Zurich, Switzerland. 1999: 144-153.[8]Sarma D D and Matula D W. Measuring the accuracy of ROM reciprocal tables. Proceedings of the 11th Symposiumon Computer Arithmetic. Windsor, Ontario, Canada. 1993: 95-102. -
计量
- 文章访问数: 3472
- HTML全文浏览量: 115
- PDF下载量: 1822
- 被引次数: 0
下载:
