D.N. Shields

A bilinear fault detection observer

A bilinear fault detection observer is proposed for a bilinear system with unknown inputs. The residual vector in the design of the observer is decoupled from the unknown inputs and, under certain conditions, is made sensitive to all the faults. Sufficient conditions are given for the existence of the observer and results are given for the explicit calculation of the observer design matrices. An application to a hydraulic drive system is given.

A bilinear fault detection observer and its application to a hydraulic drive system

A bilinear fault detection observer is proposed for bilinear systems with unknown input. A sufficient condition for the existence of the observer is given. The residual generated by this observer is decoupled from the unknown input. The method is applied to a hydraulic test rig to detect and isolate a large group of simulated faults. The effectiveness of the method in fault detection and isolation for the hydraulic system is demonstrated by real data simulation.

Design of nonlinear observers for detecting faults in hydraulic sub-sea pipelines ☆

Abstract Firstly, a general nonlaminar model is considered for pipeline dynamics, including a treatment of faults caused by pipe restrictions. For three cases results are given for stability, robustness and fault detectability of a combined observer and residual (fault detection signal). An efficient numerical design algorithm is proposed. The method is applied to an actual experimental pipeline (rig system) which is set up to model a sub-sea umbilical. Results on modelling and on observer and residual (signal) design are given. The effectiveness of the design is tested by inducing two types of fault on the rig system.

Improved SI engine modelling techniques with application to fault detection

This paper applies a fault detection strategy to a Jaguar car engine. Real data, over several engine speeds, is used to obtain an improved model of a subsystem comprising of the air intake, the manifold dynamics and the engine pumping. Modelling of an air leak is also considered. Three practical fault scenarios are assessed, deemed important by the manufacturer. A nonlinear observer method is used to detect practical sensor faults and an air leak in the manifold. Isolation logic is designed with respect to a restricted set of sensor measurements. The effectiveness of both modelling and fault detection axe assessed. A practical aim is to limit engine pollutants caused by system faults.

A bilinear fault detection filter

A bilinear fault detection filter (BFDF) is proposed for bilinear systems with unknown inputs. Sufficient conditions are given for the existence of a BFDF and a strict BFDF for which the latter is sensitive to all faults. A particular choice of observer structure is shown to give explicit solutions both for the observer gain matrix and for establishing fault sensitivity. Structured actuator and component faults give rise to residuals with known directions enabling these faults to be easily isolated. A numerical example is used to demonstrate the BFDF design.

Observer-based residual generation for fault diagnosis for non-affine non-linear polynomial systems

This paper proposes an observer-based residual generator (OBRG) for diagnosing faults in a continuous non-affine system with polynomial non-linearities up to any finite degree where the fault and an unknown input affect both the system and part of the output. Firstly, given certain assumptions and the use of defined extended vectors, a parameterized polynomial system is considerred for which a compact set of sufficient conditions is given for the existence of a candidate OBRG. Conditions for error stability (by a Lyapunov method) and detectability are given. The calculation steps in the design of the OBRG are shown to involve the solution of three linear equations (with parameterizations) and the calculation of a set of constant matrices (for detectability of faults). A result is then given establishing that the design holds for a much wider class of systems. The residual design is applied to a real three-tank system.This paper proposes an observer-based residual generator (OBRG) for diagnosing faults in a continuous non-affine system with polynomial non-linearities up to any finite degree where the fault and an unknown input affect both the system and part of the output. Firstly, given certain assumptions and the use of defined extended vectors, a parameterized polynomial system is considerred for which a compact set of sufficient conditions is given for the existence of a candidate OBRG. Conditions for error stability (by a Lyapunov method) and detectability are given. The calculation steps in the design of the OBRG are shown to involve the solution of three linear equations (with parameterizations) and the calculation of a set of constant matrices (for detectability of faults). A result is then given establishing that the design holds for a much wider class of systems. The residual design is applied to a real three-tank system.

Fault diagnosis for, a gas-fired furnace using bilinear observer method

A simulation study is undertaken on fault diagnosis for a gas-fired high temperature furnace. A single input multi output bilinear model is found to be suitable to represent the furnace. A previously proposed full-order bilinear state observer is used to generate the residual. Two actuator and two sensor faults are simulated and tested using this method.

Robust fault detection observers for nonlinear polynomial systems

This paper proposes a robust fault detection observer (RFDO) for classes of nonlinear models which consists of polynomial nonlinearities up to degree three and are structured in continuous state-space form. The observer and detection signal are designed to be independent of unknown inputs. Sufficient conditions are given, in an efficient compact form, for an RFDO (observer and signal) to exist. Several useable results on detectablility conditions are given and a numerical procedure is considered. The modelling of a hydraulic rig system is considered. Design steps are shown in obtaining an RFDO which is sensitive to faults at certain nodal points. This RFDO forms part of a bank of three observers used in a fault isolation procedure which is tested on real input-output data.

A hybrid fault diagnosis approach using neural networks

A hybrid fault diagnosis method is proposed in this paper which is based on the parity equations and neural networks. Analytical redundancy is employed by using parity equations. Neural networks then are used to maximise the signal- to- noise ratio of the residual and to isolate different faults. Effectiveness of the method is demonstrated by applying it to fault detection and isolation for a hydraulic test rig. Real data simulation shows that the sensitivity of the residual to the faults is maximised, whilst that to the unknown input is minimised. The simulated faults are successfully isolated by a bank of neural nets.

Quantitative Approaches for Fault Diagnosis Based on Bilinear Systems

Abstract fault detection and isolation (FDI) of bilinear systems is considered. Two main quantitative methods are outlined. A bilinear fault detection observer is proposed which focuses on the problem of decoupling the unknown inputs from the residuals. Secondly, a parity space method is developed for bilinear systems. Robust issues are addressed throughout the paper. Related methods are briefly discussed and two applications are considered.