Nanotech 2003 Vol. 1
Nanotech 2003 Vol. 1
Technical Proceedings of the 2003 Nanotechnology Conference and Trade Show, Volume 1

Smart MEMS and Sensor Systems Chapter 10

Microsensors, Arrays and Automatic Diagnosis of Sensor Faults

Authors: E. Gaura and R.M. Newman

Affilation: Coventry University, United Kingdom

Pages: 276 - 279

Keywords: diagnosis, microsensor, artificial intelligence

As man-made dynamical systems become increasingly complex, there is an ever-present need to ensure their safe and reliable operation [1]. These requirements extend beyond the normally accepted safety-critical systems (e.g. nuclear reactors, chemical plants, and aircraft) to new systems such as autonomous vehicles and rapid transport systems. Early detection of faults and/or malfunctions in industrial processes and systems can help reduce downtimes and the incidence of catastrophic events [1]. Sensors are essential components of any process or system which makes use of automatic control [2]. It follows that an important aspect of any process/system fault diagnosis strategy is to attempt to determine their state of functionality. During the last two decades, fault detection theory has attracted considerable academic interest and as a result, a variety of techniques for fault diagnosis have been developed. Most of these techniques are designed to work in a centralised manner, by accounting simultaneously for sensors, actuators and process/system component faults. A more recent trend is towards hierarchical health monitoring of complex processes), in which the health monitoring of sensors, for example, is considered as an individual task. The paper will explore further the possibility of local diagnosis for sensors, by developing a methodology for sensor self-diagnosis, self-validation and monitoring. It is intended to study the feasibility of using artificial neural network (ANN) techniques for implementing the above methodology. ANNs have been successfully used in a variety of applications for complex data analysis and feature extraction. In the context of the proposed research, one of their main advantages is that the majority of intensive computation takes place during the training process. Once the ANN is trained for a particular task, operation consists of propagating the data through the mapping produced by the ANN, thereby making possible real-time self-diagnosis, self-validation and monitoring. Acceleration sensors have been chosen for the case study, as their lack of accessible internal signals makes the tasks of diagnosis and validation particularly challenging. Acceleration sensors are crucial components in numerous applications, for example inertial navigation, automotive systems, avionics, vibration monitoring and control. In most such applications, accurate and reliable sensor readings are vital for good overall system performance [3,4]. Despite advances in fabrication technologies, acceleration sensors exhibit imperfections such as offset, drift, non-linearity and noise, and the magnitude of these imperfections is found to vary both from sensor to sensor and with time. Fundamental characteristics of the sensor, e.g. sensitivity, may be subject to manufacturing tolerances, varying material properties and ambient effects [2]. Moreover, during operation, as with any other system component, they may develop several types of faults and fail in a variety of ways. Several such faults will be considered and local, integrated ANN solutions for their remedy will be presented in the full paper. The justification for this research is four-fold. Firstly, localised, real-time sensor self-validation, self-diagnosis technology will enable improved automatic process/system monitoring and control. Secondly, early detection of small, incipient (rather difficult to detect) sensor faults can be achieved and therefore downtime can be reduced and catastrophes can be avoided. Thirdly, more robust, fault-tolerant control can be designed on the above basis, to accommodate/compensate for soft sensor failures (i.e., recoverable sensor failures that leave no permanent damage). Finally, easy identification of sensors which have suffered hard failures (irrecoverable sensor failures) can be achieved. References 1. Patton, R., Uppal, F., Lopez-Toribio, C. (2000). Soft computing approaches to fault diagnosis for dynamic systems: a survey, 4th IFAC Symposium on Fault Detection Supervision and Safety for Technical Processes (Safeprocess2000), Budapest, Hungary, Proceedings, pp. 298-311. 2. Gaura, E. (2000). Neural Network Techniques for the Control and Identification of Acceleration Sensors. PhD Thesis, Coventry University, UK.) 3. Alag, S., Goebel, K., Agogino, A. (1995). A framework for intelligent sensor validation, sensor fusion, and supervisory control of automated vehicles in IVHS, (ITS America) 4. Henry, M.P. (1994). Validating data from smart sensors, J. of Control Engineering, vo.41, no.9, pp. 63-66.

Microsensors, Arrays and Automatic Diagnosis of Sensor Faults

ISBN: 0-9728422-0-9
Pages: 560