基于指数神经元的模型感知器的设计

Design of Model Perceptron Based on Exponent Neuron

基于神经元感知器结构并采用最速下降自适应算法 ,针对电导、温度、湿度、水分、浓度、油粘度等测量问题设计一种基于指数神经元的自适应模型感知器 ,同时在一定意义上实现了信息融合。分析讨论了学习参数对感知器建模工作性能的影响 ,给出了电导、温度、水分应用实例的仿真曲线。结果表明 ,所设计的自适应模型感知器应用于传感器及测量系统的自动建模是有效的。该种建模方法和机制也可作为智能单元 (硬件或软件 )灵活地应用于实际测量或控制系统 。

With the structure of neuron perceptron and adaptive algorithm combined, an adaptive modeling perceptron based on exponential neuron is designed, for measurement of such measurands as temperature, moisture, humidity, conductivity, viscosity, etc. Simultaneously, in some sense, it can realize fusion of space-time signal. The effect of learning parameters upon adaptive performance is discussed. Simulation curves with adaptive algorithm are given, and the modeling perceptron proved to be effective and available. This modeling scheme and its mechanism can also be applied to actual measurement or control system as an intelligent unit(hardware or software) instead of rigid mathematical model by manual curve fitting.