Sensor selection for fault diagnostics using performance metric

Andrews, J; Reeves, J; Reeves, Jack; Remenyte-Prescott, R; Andrews, John

Abstract

As technology advances, modern systems are becoming increasingly complex, consisting of large numbers of components, and therefore large numbers of potential component failures. These component failures can result in reduced system performance, or even system failure. The system performance can be monitored using sensors, which can help to detect faults and diagnose failures present in the system. However, sensors increase the weight and cost of the system, and therefore, the number of sensors may be limited, and only the sensors that provide the most useful system information should be selected.
In this paper, a novel sensor performance metric is introduced. This performance metric is used in a sensor selection process, where the sensors are chosen based on their ability to detect faults and diagnose failures of components, as well as the effect the component failures have on system performance. The proposed performance metric is a suitable solution for the selection of sensors for fault diagnostics. In order to model the outputs that would be measured by the sensors, a Bayesian Belief Network (BBN) is developed. Sensors are selected using the performance metric, and sensor readings can be introduced in the BBN. The results of the BBN can then be used to rank the component failures in order of likelihood of causing the sensor readings. To illustrate the proposed approach, a simple flow system is used in this paper.

Citation

Andrews, J., Reeves, J., Reeves, J., Remenyte-Prescott, R., & Andrews, J. (2018). Sensor selection for fault diagnostics using performance metric. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability, 233(4), 537-552. https://doi.org/10.1177/1748006X18804690

• Journal Article Type: Article
• Acceptance Date: Sep 10, 2018
• Online Publication Date: Oct 10, 2018
• Publication Date: Oct 10, 2018
• Deposit Date: Sep 20, 2018
• Publicly Available Date: Sep 20, 2018
• Journal: Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability
• Print ISSN: 1748-006X
• Electronic ISSN: 1748-0078
• Publisher: SAGE Publications
• Peer Reviewed: Peer Reviewed
• Volume: 233
• Issue: 4
• Pages: 537-552
• DOI: https://doi.org/10.1177/1748006X18804690
• Keywords: Sensor selection; Bayesian Belief Network; Fault diagnostics
• Public URL: https://nottingham-repository.worktribe.com/output/1094256
• Publisher URL: http://journals.sagepub.com/doi/10.1177/1748006X18804690
• Contract Date: Sep 20, 2018