Search

A new generation of power electronic semiconductor devices are being developed for the benefit of space and terrestrial harsh-environment applications. 200-600 V lateral transistors and diodes are being fabricated in a thin layer of silicon (Si) wafer bonded to silicon carbide (SiC). This novel silicon-on-silicon-carbide (Si/SiC) substrate solution promises to combine the benefits of silicon-on-insulator (SOI) technology (i.e device confinement, radiation tolerance, high and low temperature performance) with that of SiC (i.e. high thermal conductivity, radiation hardness, high temperature performance). Details of a process are given that produces thin films of silicon 1, 2 and 5 μm thick on semi-insulating 4H-SiC. Simulations of the hybrid Si/SiC substrate show that the high thermal conductivity of the SiC offers a junction-to-case temperature ca. 4× less that an equivalent SOI device; reducing the effects of self-heating, and allowing much greater power density. Extensive electrical simulations are used to optimise a 600 V laterally diffused metal-oxide-semiconductor field-effect transistor (LDMOSFET) implemented entirely within the silicon thin film, and highlight the differences between Si/SiC and SOI solutions.

As sinusoidal inputs are common in engineering systems, it is therefore important to develop fault detection methods for these systems in order to improve their reliability. In this paper, a fault detection method for nonlinear systems with multitone sinusoidal input is developed. By assuming the output of the system has the almost periodicity property, the proposed method is derived based on the periodogram for fault detection using the asymptotic local approach, which is effective to detect both incipient and abrupt faults. It is shown that the fault detection problem using the proposed method is reduced to a [chi square]-test for a given confidence level. Simulation example involving a nonlinear system is used to illustrate the implementation and the performance of the proposed method. Index Terms--Asymptotic local approach, fault detection, periodogram, sinusoidal signal, [chi square]-test.

This paper studies the bounds of the response of a class of nonlinear systems with sinusoidal inputs. The unknown nonlinear system is assumed to have a unique steady-state response for a given input under different initial conditions. The sufficient condition of this assumption is given, and the bounds of the transient response and the steady-state response are derived in the sense of 2-norm and [L.sup.2] norm, respectively. A simulation example is presented to demonstrate the results. Index Terms--Sinusoidal input, steady-state response, transient response, bound.

A novel fault detection (FD) method for nonlinear systems using the residuals generated by the unscented Kalman filter (UKF) is proposed. The errors of the UKF are derived and sufficient conditions for the convergence of the UKF are presented. As the local approach is a powerful statistical technique for detecting changes in the mean of a Gaussian process, it is used to devise a hypothesis test to detect faults from residuals obtained from the UKF. Further, it is demonstrated that the selection of a sample number is important in improving the performance of the local approach. To illustrate the implementation and performance of the proposed technique, it is applied to detect sensor faults in the measurement of satellite attitude.

Reliable control of superheated steam temperature is necessary to ensure high efficiency and high load-following capability in the operation of modern power plant. This is often difficult to achieve using conventional PI controllers, as power plants are nonlinear and contain many uncertainties. A nonlinear generalized predictive controller based on neuro-fuzzy network (NFGPC) is proposed in this paper, which consists of local GPCs designed using the local linear models of the neuro-fuzzy network that models the plant. The proposed nonlinear controller is applied to control the superheated steam temperature of a 200-MW power plant. From the experiments on the plant and the simulation of the plant, much better performance than the traditional cascade PI controller or the linear GPC is obtained. Index Terms--Generalized predictive control, neuro-fuzzy networks, superheated steam temperature.

A multiple fault detection and identification method based on fully decoupled parity equations for dynamic systems with known linear and unknown nonlinear terms is presented. The fully decoupled parity equation vectors is derived and it is shown that the residuals generated from it are decoupled from other faults and the unknown nonlinear term and are sensitive only to specific actuator or sensor faults. The condition for the existence of the equation is also given. From the residuals generated from the fully decoupled parity equation, the faults are estimated using the recursive least-squares method. The performance of the proposed method is illustrated by applying it to detect, isolate, and identify faults in a simulated dc motor. Index Terms--Fault detection and identification, parity equation, recursive least-squares method.