用神经网络求解P－t可诊断模型下的计算机系统诊断问题

Computer System Diagnosis Problem under P-t Diagnosable Model Solved by A Artificial Neural Network

系统地研究了如何构造一个Ｔａｎｋ＆Ｈｏｐｆｉｅｌｄ型神经优化网络来求解计算机系统诊断的Ｐ－ｔ可诊断模型的诊断问题。模拟实验表明神经网络对初始输入和连接权均具有容错性；模拟实验得出了电容Ｃ。

The problems were systematically studeid of how to form a Tank& Hopfield-type neural optimization[4] network to solve the diagnosis for P-t diagnosiable model of computer di-agnosis. The simulation seems effective to solve the computer system diagnosis problems by neuralnetworks. The simulation also shows that the neural networks are of the character of fault toleranceto initial input and connection weights,i.e.their slight deviations do not offect the propor functionsof the network; the simulation reveals the relations among capacity C、resistance R、gain K and time.