In a large-scale system such as a power plant which is strongly nonlinear, it is very difficult to locate faults origins when enormous abnormal signals come into exitence in the limited time due to the plant complexit...
详细信息
In a large-scale system such as a power plant which is strongly nonlinear, it is very difficult to locate faults origins when enormous abnormal signals come into exitence in the limited time due to the plant complexity. In this paper, a new approach is proposed in which hierarchically structured neural networks are utilized based on the system directed graph (digraph) with signed gain branches, and a simple method of training data extraction is also given for faults detection, which is crucial aspect for real-time diagnosis in a complex system. This method also comprises the fault propagation probability, fault propagation time, and the fault rates of devices for more accurate diagnosis, and provides auto-tuning of the states of SDG nodes according to the various operating conditions through the plant structure identifier. For each subsystem, there corresponds neural network that performs the faults detection, and the estimation of their faults size. The plant-wise faults diagnosis is performed for removing the spurious faults using the information from the lower level neural networks, and provides the correponding treatment guideline to the operator at the top level neural network. The information on the faults size can give the plant operators the plant-wise control strategy. In a case study of applicaion to a pump system, the proposed scheme diagnoses multi-fault as well as a single fault including sensor fault itself in real-time during quasi-steady state condition.
Quantitative Feedback Theory (QFT) is applied to design flight control laws for the AlAA controls Design Challenge nonlinear aircraft model. The control laws are designed in two steps. First, the aircraft model is lin...
详细信息
It is demonstrated how a neural network may be used to generate a time-suboptimal control policy in continuous time. Approximation of optimal feedback by neural networks is proposed. Training data are taken from a set...
详细信息
It is demonstrated how a neural network may be used to generate a time-suboptimal control policy in continuous time. Approximation of optimal feedback by neural networks is proposed. Training data are taken from a set of precalculated open-loop trajectories from which a neural network extracts the information about the mapping in question. Simulation results are provided for a simple example.< >
A closed-loop feedback scheme for obtaining a goal microstructure during Hot Isostatic Pressing (HIP'ing) of powders is described. The control scheme relies on previously developed process models describing the pr...
详细信息
Artificial feedforward networks are studied as nonlinear function approximators used to identify forward and inverse mappings of discrete time dynamic systems. They are found to provide significant advantages over oth...
详细信息
Artificial feedforward networks are studied as nonlinear function approximators used to identify forward and inverse mappings of discrete time dynamic systems. They are found to provide significant advantages over other modelling techniques such as polynomial approximations, especially if the extrapolation beyond the region covered by the learning data is involved. We apply the neural network methodology to a simple second order approximation of a single-machine infinite-bus power system controlled by means of modifying the reactance of the line. Accurate off-line identification of forward and inverse dynamics of the system is performed by means of single hidden layer neural networks, and both models are then used in a direct inverse control configuration. The controller simulations show very good quality of transients for severe short-circuit fault.
The authors study H/sup infinity /-optimal control of singularly perturbed linear systems under general imperfect state measurements using infinite-horizon formulations. Using a differential game theoretic approach, t...
详细信息
The authors study H/sup infinity /-optimal control of singularly perturbed linear systems under general imperfect state measurements using infinite-horizon formulations. Using a differential game theoretic approach, they first show that, as the singular perturbation parameter in approaches zero, the optimal disturbance attenuation level for the full-order system under a quadratic performance index converges to the maximum of the optimal disturbance attenuation levels for the slow and fast subsystems under appropriate slow and fast quadratic cost functions. Then, they construct a controller based on the slow subsystem only, and obtain conditions under which it delivers a desired performance level even though the fast dynamics have been completely neglected. The ultimate performance level achieved by this slow controller can be uniformly improved by a composite controller that uses some feedback from the output of the fast subsystem. The authors construct one such controller, using a two-step sequential procedure, which uses static feedback from the fast output and dynamic feedback from an appropriate slow output, each obtained by solving appropriate in -independent lower-dimensional H/sup infinity /-optimal control problems.< >
We study the H ∞ -optimal control of singularly perturbed linear systems under perfect state measurements. Using a differential game theoretic approach, we show that as the singular perturbation parameter ϵ approache...
详细信息
We study the H ∞ -optimal control of singularly perturbed linear systems under perfect state measurements. Using a differential game theoretic approach, we show that as the singular perturbation parameter ϵ approaches zero, the optimal disturbance attenuation level for the full-order system under a quadratic performance index converges to a value that is bounded above by the maximum of the optimal disturbance attenuation levels for the slow and fast subsystems under appropriate "slow" and "fast" quadratic cost functions. Furthermore, we construct a composite controller based on the solution of the slow and fast games, which guarantees a desired achievable performance level for the full-order plant, as ϵ approaches zero. A "slow" controller, however, is not generally robust in this sense, but still under some conditions, which are delineated in the paper, the fast dynamics can be totally ignored. The paper also studies optimality when the controller includes a feedforward term in the disturbance.
An iteration method is presented for determining the largest singular value (2-norm) of a matrix, and its corresponding singular vectors. Connections with the power method and Bernoulli's method are presented. A f...
详细信息
An iteration method is presented for determining the largest singular value (2-norm) of a matrix, and its corresponding singular vectors. Connections with the power method and Bernoulli's method are presented. A formula is derived which describes the relationship between a matrix perturbation and the perturbation of its singular values.< >
A number of robust stability problems take the following form: A polynomial has real coefficients wvhich are multiaffine in real parameters that are confined to a box in parameter space. An efficient method is require...
详细信息
A number of robust stability problems take the following form: A polynomial has real coefficients wvhich are multiaffine in real parameters that are confined to a box in parameter space. An efficient method is required for checking the stability of this set of polynomials. We present two sufficient conditions in this paper. They involve: checking certain properties at the corners and edges of the parameter space box.
A closed-loop feedback scheme for obtaining a goal microstructure during hot isostatic pressing of powders is described. The control scheme relies on previously developed process models describing the process dynamics...
详细信息
A closed-loop feedback scheme for obtaining a goal microstructure during hot isostatic pressing of powders is described. The control scheme relies on previously developed process models describing the process dynamics during a pressing run and sensors which can measure density and grain size. Constantly updated linearization and coprime factorization are used, so the control can be implemented by convex programming. Simulation results showing the performance of the control scheme are presented.< >
暂无评论