基于BP 神經網絡能以任意精度逼近任何非線性連續函數的原理。通過在MATLAB
環境下，對典型的不穩定、非線性、強耦合的倒立擺系統建立了BP 神經網絡辨識結構，并對辨識結果進行了分析。結果表明BP 神經網絡對非線性倒立擺模型的辨識是有效的。
關鍵詞：BP 神經網絡；倒立擺 ；系統辨識
BP Neural Network for Non-linear Inverted Pendulum System Identification YAN xin ，ZHAO ying kai，DOU dong yang (College of Automation,NanJing Univ. of Technology，JiangSu 210009 ,China) Abstract:Base on the theory that the BP neural network can compel any non-linear function by any precision. Under the circumstance of MATLAB, the neural network structure identification of a typical unstable, non-linear and strong coupling inverted pendulum system is established. Analyze of the result of identification prove that it is effective for neural network to identify the non-linear inverted pendulum system.
Key words: BP Neural Network ；Inverted Pendulum；System Identification