基于神经网络的电子节气门系统模型参考自适应控制
|
摘要
使用多层感知器神经网络模型来识别和控制非线性电子节气门系统。首先,神经网络模型在不同运行条件下辨识,它代表非线性节气门伺服系统的动态特性。其次,使用油门辨识器网络模型来设计和训练神经网络控制器模型,从而使节气门系统的追踪控制位置遵循参考模型。油门辨识器网络模型用于辅助以离线模式训练的神经网络控制器。神经网络控制器使用相同的输入来进行训练,这些输入被反馈到实际的节气门系统以产生相同的输出。通过调整神经网络控制器的权重和偏差参数,使用自适应算法来减小输出之间的差异。对使用神经网络控制器的节气门控制系统的跟踪控制性能与使用经典自适应PID控制器进行比较。仿真结果表明:采用神经网络控制器可实现跟踪控制,满足控制性能的所有需求。
Having multi-layer perception model of the neural network adopted to identify and control a non-linear electronic throttle system was implemented. Firstly, the neural network model which represents dynamic behavior of the non-linear throttle servo system can identify the non-linear electronic throttle system at different operating conditions; and secondly, the throttle identifier network model can be used to design and train the neural network controller model so that the tracking control position of the throttle system follows a reference model. Having the throttle identifier network model used to assist the neural controller trained in off-line mode was implemented and the neural network controller trained with the same inputs which are fedback to the actual throttle system to produce the same output. Through adjusting the weights and deviation parameters of the controller network model, an adaptation algorithm was used to reduce the difference between the outputs. Comparing the tracking control performance of the throttle control system which using the neural network controller with the classical adaptive PID controller shows that, making use of the neural network controller can realize the tracking control and satisfy all requirements for the control performance.
Having multi-layer perception model of the neural network adopted to identify and control a non-linear electronic throttle system was implemented. Firstly, the neural network model which represents dynamic behavior of the non-linear throttle servo system can identify the non-linear electronic throttle system at different operating conditions; and secondly, the throttle identifier network model can be used to design and train the neural network controller model so that the tracking control position of the throttle system follows a reference model. Having the throttle identifier network model used to assist the neural controller trained in off-line mode was implemented and the neural network controller trained with the same inputs which are fedback to the actual throttle system to produce the same output. Through adjusting the weights and deviation parameters of the controller network model, an adaptation algorithm was used to reduce the difference between the outputs. Comparing the tracking control performance of the throttle control system which using the neural network controller with the classical adaptive PID controller shows that, making use of the neural network controller can realize the tracking control and satisfy all requirements for the control performance.
引文
[1] 李建兵,李新华.基于神经网络的电子节气门开启角度的模拟研究[J].湖南农机,2008,(5):47～48.
[2] 黄永刚.电子节气门系统建模与先进控制方法研究[D].南京:东南大学,2017.
[3] 李建磊.基于湿式自动离合器的商用车起步控制策略仿真研究[D].长春:吉林大学,2017.
[4] Yang H,Liu L Y,Wasacz B.Neural Network Based Feedforward Control for Electronic Throttles[C].SAE2002 World Congress.Warrendale,PA:Society of Automotive Engineers,2002:1141～1149.
[5] Zribi M,Al-Zamel A.Field-Weakening Nonlinear Control of a Separately Excited DC Motor[C].First International Symposium on Control,Communications and Signal Processing.Piscataway,NJ:IEEE,2004:1～22.
[6] De Jesus O,Hagan M T.Backpropagation through Time for a General Class of Recurrent Network[C].Proceedings of the IEEE International Joint Conference on Neural Networks.Piscataway,NJ:IEEE,2001:2638～2643.
[2] 黄永刚.电子节气门系统建模与先进控制方法研究[D].南京:东南大学,2017.
[3] 李建磊.基于湿式自动离合器的商用车起步控制策略仿真研究[D].长春:吉林大学,2017.
[4] Yang H,Liu L Y,Wasacz B.Neural Network Based Feedforward Control for Electronic Throttles[C].SAE2002 World Congress.Warrendale,PA:Society of Automotive Engineers,2002:1141～1149.
[5] Zribi M,Al-Zamel A.Field-Weakening Nonlinear Control of a Separately Excited DC Motor[C].First International Symposium on Control,Communications and Signal Processing.Piscataway,NJ:IEEE,2004:1～22.
[6] De Jesus O,Hagan M T.Backpropagation through Time for a General Class of Recurrent Network[C].Proceedings of the IEEE International Joint Conference on Neural Networks.Piscataway,NJ:IEEE,2001:2638～2643.