竞争学习的非均匀采样非线性系统的模糊辨识

在实际非线性系统中,由于资源的限制,使得输入信号快速刷新,输出信号慢速采样.利用获得的非均匀采样数据对原非线性系统辨识存在一定困难.为此,通过提升技术,把非线性系统的多个特征点局部的线性模型转化为模糊模型的后件线性模型.在此基础上,提出基于竞争学习和递推梯度下降方法的辨识算法.通过定理证明:输入信号在持续激励条件下,模糊模型的参数能够一致性收敛;针对化工p H中和过程非线性系统,采用非均匀采样数据,建立其模糊模型,通过实际数据与模糊模型输出数据误差对比,表明了实际系统在非均匀采样条件下,模糊辨识能够建立其过程模型,验证了提出方法的有效性.

Fuzzy identification of non-uniformly multirate sampled nonlinear systems based on competitive learning

In practical nonlinear system, due to the limitation of resources, the input signal is quickly refreshed, while the output signal is slowly sampled. Thus, it is difficult to identify the original nonlinear system by using the sampled data. For this purpose, the linear models of multiple characteristic points of nonlinear system are transformed into a series of consequent linear models of the fuzzy model by the lifting technique. On this basis, we propose a fuzzy identification algorithm based on competitive learning and recursive gradient descent method. And we prove that the parameters of the fuzzy model can be uniformly convergent under the condition of persistent excitation. In view of chemical pH neutralization process, the fuzzy model of the chemical system is established by using non-uniformly sampled data. By comparing the output errors between the actual data and the output data of the fuzzy model, it is shown that the fuzzy identification method can establish the process model in the real system under the condition of non-uniform sampling, which verifies the validity of the proposed method.