This paper compares radial basis function networks for identification of nonlinear dynamic systems with classical methods derived from the Volterra series. The performance of these different approaches, such as Hammer...
详细信息
This paper compares radial basis function networks for identification of nonlinear dynamic systems with classical methods derived from the Volterra series. The performance of these different approaches, such as Hammerstein, Wiener and NDE models, is analysed. Since the centres and variances of the Gaussian radial basis functions will be fixed before learning and only the weights are learned, a linear optimization problem arises. Therefore training the network and parameter estimation becomes comparable in computational effort. It is shown that the classical methods can compete or even perform better than the neural network, if the assumptions for the structure are valid. However, in practical applications when the structure is not known the radial basis function network performs much better than the classical methods.
This paper considers the application of neural networks with distributed dynamics to the identification of nonlinear systems. The primary objective is to establish a neural model bank which generates prediction errors...
详细信息
This paper considers the application of neural networks with distributed dynamics to the identification of nonlinear systems. The primary objective is to establish a neural model bank which generates prediction errors. These estimation residuals can be treated as analytic symptoms to supervise the plant operating state. The practical approach applicability has been illustrated using a thermal plant. Here, two sensor faults of interest are localized by means of the so-called residual pattern.
Identification of linear and nonlinear dynamic systems can be performed with a series-parallel or parallel model. In this paper both approaches are compared. While many applications require the model to run in paralle...
详细信息
Identification of linear and nonlinear dynamic systems can be performed with a series-parallel or parallel model. In this paper both approaches are compared. While many applications require the model to run in parallel to the process, usually the identification procedure is carried out with a series-parallel model. This paper shows that optimization of a series-parallel model does not necessarily lead to a good parallel model. Furthermore a decrease of the error in one configuration may result in an increase in the other.
The dynamical engine test stand described with a real engine and a drive-line simulation is a capable tool for exhaust emission analysis. To simulate the dynamics of load changes occurring in a driving vehicle, a mode...
详细信息
The dynamical engine test stand described with a real engine and a drive-line simulation is a capable tool for exhaust emission analysis. To simulate the dynamics of load changes occurring in a driving vehicle, a model of the car body including clutch, manual gear-box, drive-line, differential and wheels has been developed and implemented on the test stand. Exhaust emission tests may be performed automatically with a proper driver simulation used in the dynamical engine test stands. Starting vehicle from standstill is a critical operation, especially for vehicles with manual gearbox, that occurs frequently in exhaust emission test cycles. Model-based control strategies for the vehicle startup have been developed. These strategies are presented and discussed.
This paper compares radial basis function networks for identification of nonlinear dynamic systems with classical methods derived from the Volterra series. The performance of these different approaches as Hammerstein,...
详细信息
This paper compares radial basis function networks for identification of nonlinear dynamic systems with classical methods derived from the Volterra series. The performance of these different approaches as Hammerstein, Wiener and NDE models is analysed. Since the centres and variances of the Gaussian radial basis functions will be fixed before learning and only the weights are learned, a linear optimization problem arises. Therefore training the network and parameter estimation become comparable in calculational effort. It will be shown, that the classical methods can compete or even perform better than the neural network, if the assumptions for the structure are valid. However if in practical applications the structure is not known the radial basis function network performs much better than the classical methods.
In this contribution several approaches to nonlinear adaptive control of mechanical drives are proposed taking into account determined nonlinear actuator characteristics such as friction, hysteresis and other static n...
详细信息
In this contribution several approaches to nonlinear adaptive control of mechanical drives are proposed taking into account determined nonlinear actuator characteristics such as friction, hysteresis and other static nonlinearities. Improved control performance in terms of dynamics, high positioning accuracy and robustness is achieved by combining linear position control with nonlinear feed-forward and adaptive friction compensation methods. Based on a precise model of the drive dynamics recursive identification techniques are applied to obtain reliable estimates of the time varying and position dependent process parameters, the nonlinear friction characteristic in particular. As examples the nonlinear friction characteristics of robotic joint drives and pneumatic drives are determined and it is shown how the control can be improved considerably by model based nonlinear control in the presence of time varying nonlinear characteristics and hysteresis effects. Experimental results illustrate the efficiency of the developed nonlinear control schemes.
Many systems with nonlinear behaviour are difficult to model. Consequently model based approaches to fault detection can not be applied easily. Additional problems are introduced in closed loops by the controversity o...
详细信息
Many systems with nonlinear behaviour are difficult to model. Consequently model based approaches to fault detection can not be applied easily. Additional problems are introduced in closed loops by the controversity of controller design and fault detection. For a certain class of processes with mainly constant input signals a combination of parameter estimation techniques and the parity space approach allows on-line fault detection with small signal models. Only few knowledge of the process is required: An idea about the dynamic model order and the static nonlinear characteristic curve. Using the proposed method fault detection is performed for temperature control loops in a thermic plant.
A methodology for designing an automated vehicle longitudinal velocity and distance controller is presented and applied to an automobile. nonlinearities. The controller consists of a three layer structure. In the firs...
详细信息
A methodology for designing an automated vehicle longitudinal velocity and distance controller is presented and applied to an automobile. nonlinearities. The controller consists of a three layer structure. In the first layer a linearization of the nonlinearities is done in order to achieve a simplified structure for controller design purposes. With respect to changes in the typical vehicle parameters-mass and aerodynamical drag-an adaptive controller structure is used and these parameters are estimated by a recursive least squares algorithm. Based on classical controlling techniques a linear acceleration controller is developed in the middle layer. A fuzzy controller is applied in the upper layer. This controller based on the linguistic description of comfort demands. Additionally a neural network is used in this layer instead of the fuzzy system. The complete structure is used in two different series fabricated vehicles and experimental results are shown for highway traffic and also for stop-go traffic on highway congestions. Additionally a neural network is trained by measurement data as a second approach to describe the comfort demands.
A methodology for designing an automated vehicle longitudinal velocity and distance controller is presented and applied to an automobile. Typically a vehicle is described by velocity depended dynamics and specific non...
详细信息
A methodology for designing an automated vehicle longitudinal velocity and distance controller is presented and applied to an automobile. Typically a vehicle is described by velocity depended dynamics and specific nonlinearities. Therefore the controller consists of a three layer structure. In the first layer a linearization of the nonlinearities is done in order to achieve a simplified structure for controller design purposes. With respect to changes in the typical vehicle parameters-mass and aerodynamical drag - an adaptive controller structure is used and these parameters are estimated by a recursive least squares algorithm. Based on classical controlling techniques a linear acceleration controller is developed in the middle layer. Up to now no mathematical equations are available describing the subjective demands of the driver and passengers during longitudinal motion. Therefore a fuzzy controller is applied in the upper layer. This controller based on the linguistic description of comfort demands. The complete structure is used in two different series fabricated vehicles and experimental results are shown for highway traffic as well as for stop-go traffic on highway congestions.
An approach for supervision of vehicle dynamics is presented which may be used for intelligent vehicle control and state monitoring. In particular, the task is to determine automatically an altered road surface for ve...
详细信息
An approach for supervision of vehicle dynamics is presented which may be used for intelligent vehicle control and state monitoring. In particular, the task is to determine automatically an altered road surface for vehicles (may be due to weather conditions) with possibly lower adhesion between tires and road. Fault detection methods like the parity space approach can be used to detect changes in technical processes. These changes indicate different process behaviour in a new operation range. This may be of special importance when the new behaviour is more dangerous for the driver and the passengers of a car. In a further step, suitable classification algorithms were used to detect the actual driving state by processing the residual output.
暂无评论