SOURCE:

Faculty: Engineering
Department: Electronic And Computer Engineering

CONTRIBUTORS:

Ofoegbu, O. E.
Inyiama, H. C.

ABSTRACT:

A Fault Detection and Isolation (FDI) data acquisition system for gaseous pollutants monitoring was developed in this dissertation. Missing measurement data due to sensor drop out faults, error in data measurement due to offset bias faults and variance degradation faults in a data acquisition system all pose serious challenges to overall system reliability when in a deployment. Thus a fault tolerant approach using redundant sensors and a supervisory controller deployed in a multi-tier data acquisition system architecture should detect the occurrence of these faults and have capability to isolate/accommodate the occurring fault. Additive and subtractive offset bias faults in data acquisition systems are majorly caused by environmental factors, while variance degradation faults are caused by ageing in sensors, which if unattended to would result in a drop out sensor fault. Thus, using a redundant sensor, gives room for switching between sensors in a given node when one gets old, and with the supervisory controller controlling environmental factors which could cause an offset in sensor reading. The FDI data acquisition system has a multi-tier architecture comprising of a data acquisition system unit, a central remote terminal unit and a data analytics unit. An FDI mathematical model based on data variance and deviation was implemented in the central terminal unit to detect the occurring faults, while feedback information of these faults was sent back to the data acquisition unit, whose fault isolation/accommodation behavior was modeled using an error residual generator. An Air Quality Monitoring System application (AQMS) provided the tool for the data analytics phase of the study. The AQMS application allowed for trend analysis, where results gotten from pre-analysis by the central remote terminal unit and post –analysis of the supervisory controller using the developed FDI, showed that CO and NO measurements of (40, 120) and (187,110) received on the 12/09/2014, were shown to be below the CO sensor lower threshold of 75mg/m3 and above the NO sensor upper threshold of 107mg/m3, thus indicating offset bias faults and data variance faults respectively. Data received afterwards didn’t indicate a continuous occurrence of the fault situation. Data trend analysis for multiple hours, revealed a change in concentration of CO by values<±16.64mg/m3 and NO by values< ±9.14mg/m3, where values outside this range indicate a fault situation. This research offered a simplistic approach to fault tolerant deployment with fewer components as opposed to the popular triple modular redundancy (TMR) methods. The new system can be applied to any fixed process application for pollution study of gaseous pollutants as it offers a more reliable operation than a generic data acquisition system.