论文部分内容阅读
基于人工神经网络气体分析原理与误差形成原因的分析基础上，指出样本质量与网络学习的局部最小是影响气体分析精度的主要原因。研究和建立了样本预处理系统和网络的改进学习算法，对二元混合气的分析测试进行了试验。结果显示，上述措施有效改善和提高了神经网络对气体分析的精度。
Based on the analysis of the principle of gas analysis and the cause of error formation, it is pointed out that the sample quality and the local minimum of network learning are the main reasons that affect the gas analysis accuracy. The improved learning algorithm of sample preprocessing system and network was researched and established, and the analysis of binary mixed gas was tested. The results show that the above measures effectively improve and improve the accuracy of neural network for gas analysis.