目的：将人工神经网络用于安痛定注射液紫外光谱的定量分析。方法：网络结构为三层，输入节点数为10，输出节点数为2，隐含节点数为5。算法为Levenberg-marqardt优化方法。结果：氨基比林和安替比林的平均回收率分别为99.2%和100.3%，相对标准差分别为1.9%和1.7%(n=9)。与地方标准方法比较，两者测定结果无显著差异。结论：本法可用于安痛定注射液中氨基比林和安替比林的测定。 OBJECTIVE: To apply artificial neural network to the quantitative analysis of the ultraviolet spectrum of Antongding injection. Methods: The network structure is three layers, with 10 input nodes, 2 output nodes and 5 hidden nodes. The algorithm is the Levenberg-marqardt optimization method. Results: The average recoveries of aminopyrine and antipyrine were 99.2% and 100.3%, respectively, with relative standard deviations of 1.9% and 1.7%, respectively (n = 9). Compared with the local standard method, there is no significant difference between the two methods. Conclusion: This method can be used for the determination of aminopyrine and antipyrine in Antongding Injection.