CAJ | 학술논문

针对复杂非线性动态系统辨识问题,提出了一种基于过程神经元网络(PNN)的辨识模型和方法.根据系统待辨识的模型结构和反映系统模态变化特征的动态样本数据,利用PNN对时变输入/输出信号的非线性变换机制和自适应学习能力,建立基于PNN的系统辨识模型.辨识模型能够同时反映多输入时变信号的空间加权聚合以及阶段时间效应累积结果,直接实现非线性系统输入/输出之间的动态映射关系.文中构建了用于并联结构和串一并联结构辨识的PNN模型,给出了相应的学习算法和实现机制,实验结果验证了模型和算法的有效性.
침대복잡비선성동태계통변식문제,제출료일충기우과정신경원망락(PNN)적변식모형화방법.근거계통대변식적모형결구화반영계통모태변화특정적동태양본수거,이용PNN대시변수입/수출신호적비선성변환궤제화자괄응학습능력,건립기우PNN적계통변식모형.변식모형능구동시반영다수입시변신호적공간가권취합이급계단시간효응루적결과,직접실현비선성계통수입/수출지간적동태영사관계.문중구건료용우병련결구화천일병련결구변식적PNN모형,급출료상응적학습산법화실현궤제,실험결과험증료모형화산법적유효성.
Aiming at the identification of complex nonlinear dynamic system, an identification model and method based on process neural network (PNN) is proposed. According to the model structure which is to be identified and the dynamic sample data which reflect system modal verification characteristics, a system identification model based on PNN is set up using nonlinear transform mechanism and self-adaptive learning ability of PNN to the relationship between time-varying input signals and output signals. The identification model can reflect spatial weighted aggregation and time effect accumulation result to multi-input time-varying signals at the same time, and the dynamic input-output mapping relationship of nonlinear system can be found directly. A PNN model for parallel structure and serial-parallel structure is constructed, and the corresponding learning algorithm and realization mechanism are given. The experiment results verify the effectiveness of the model and algorithm.