用神经网络算法预测氢蚀孕育期

ARTIFICIAL NEURAL NETWORK TECHNOLOGY FOR THE DATA PROCESSING OF HYDROGEN ATTACK

利用神经网络的分析方法,对高温氢腐蚀进行分析建模,综合考虑温度、氢分压和氢腐蚀孕育期之间的关系,为准确的数学建模、预测设备的使用寿命和相应专家系统的研制开拓了新的思路．

Artificial neural networks are forms of artificial intelligence that learn correlative patterns between input and output information without a specific model. Then they use the learned relationships to make predictions. Five back - propagation artificial neural networks were constructed to recognize certain relationships in hydrogen attack to predict the time to incubate fissur-ing of 0. 5Mo and carbon steel. Given the two parameters, environmental temperature and hydrogen pressure, these neural network models can offer a prediction with certain accuracy. The feasibility of the model was verified by the data from papers. It is proved that the ANN technology is applicable to the evaluation of the complicated system of hydrogen attack, and also gives a base to develop the expert system and residual life evaluation system for hydrogen attack.