Rahmatallah Shoureshi

Failure Detection and Isolation using System Structure Knowledge and Eigenvalue Sensitivity

This paper presents the development of a model reference innovations based failure detection and isolation (FDI) scheme. Recently, there has been many researches carried out on failure detection schemes. However, isolation and identification of the failed components is a very important step in FDI and this paper emphasizes this aspect. Eigenvalue sensitivity analysis is introduced to develop an apriori rating for failed component isolation. This rating determines which failures are more influential to system behavior. Application of this FDI scheme to a simple fuel flow control system illustrates its feasibility on failure detection and how failed components can be ranked from least probable to most probable using eigenvalue sensitivity.

An Optically Driven Learning Control for Industrial Manipulators

Dynamic interactions between the robot manipulator and its environment introduce control complications, and unmodeled dynamics and flexibilities in the manipulator itself introduce additional uncertainties. An optically based learning controller that can utilize past experience and sensory information about the current state to overcome these difficulties is presented. An inexpensive, single-function optical sensor uses a light-emitting diode and eight detectors to provide information on the distance normal to the surface of a workpiece. The method is implemented on a GE-P50 robot and can be used to track the unknown surface of the workpiece.<<ETX>>

Analytical and Experimental Investigation of Flow-Reversible Heat Exchangers using Temperature-Entropy Bond Graphs

Modeling of heat exchangers using true bond graphs with temperature and rate of change of entropy as power variables is presented. Techniques used for modeling of irreversabilities and compressible flows are shown. The results of two and three lump models are compared with experimental results, with the agreement between these low order models and the experimental results being good. This paper shows how well a three lump model (6th order) can predict the dynamics of an actual reversal of flow. Heat exchanger response to mass flow rate oscillation is presented.

Dynamic modelling of coupled electroacoustic systems using bond graph

One of the main advantages of bond graph over other dynamic modeling techniques is its unified approach toward coupled multi-en ergy doma in syste ms. The pa per presents application of bond graphs...

A Robust Failure Diagnostics Scheme for Nonlinear Thermofluid Processes

The catastrophic failure of many physical systems is often precipitated by degradation of minor system components. A real time microprocessor based diagnostic scheme may detect system abnormalities and apply various strategies to determine whether compensatory action is required. In this paper, a robust innovations-based failure detection methodology for improved heat pump performance is presented. A sixth-order nonlinear model and an extended Kalman filter were used to reconstruct the systems state based upon limited sensor feedback. The estimated mass flow rates and pressures were statistically monitored and failures predicted based upon the repeated violation of an alarm threshold. A series of heat pump failures are experimentally induced and representative results presented and analyzed to demonstrate the potential advantages of this diagnostic approach.

Application of Bond Graph to Thermofluid Processes and Systems

Over the past few years a study was focused on development of bond graphs for thermofluid processes and systems using the true power variables of temperature and time rate of change of entropy. This paper summarizes the steps taken in this study. Discussion begins with study of a simple case of single phase incompressible flow and ends with a completely general case of multi-phase, variable density flow. Variations in density require introduction of momentum equation to the bond graph. Inclusion of entropy as a state variable necessitates formulation of Gibbs equation and its representation by means of bond graphs. This paper presents these formulations and representations, and compares dynamic results predicted by bond graphs with classical approaches.