Reduction of No Fault Found (NFF) through improved design

In 2011, I held a Postdoctoral researcher position at Cranfield University on an EPSRC funded project. The research was focused on developing fault detection and isolation algorithms within military aerospace applications. This research opportunity enabled me to continue my research into fault detection coupled with design and process optimisation within engineering. It helped me grow to have more mature outlook at my research work as the project required both high standards of programming together with extensive control knowledge. Over the four years, this work enabled me to develop algorithms for improved on-board diagnostics and sensor placement optimisation whilst developing models for using built-in-tests and adaptive control. A number of artificial intelligence techniques were developed for on-board diagnostic systems that provide the system controller with a ‘snapshot’ of its condition and allows it to make performance decisions for the overall system. This work help develop technology for various electro-mechanical systems in UAVs. I have also made use of my PhD experience and developed non-uniform sampling concepts to investigate intermittent fault behaviour and have been able to publish my work in high impact journals. I also acquired advance knowledge of computational tools (R, Python) to develop my algorithms and analyse data. Below is a summary of my various research efforts:

1) Predicting of faults in avionic units: I focused on identifying equipment design features (i.e. functions/components, measures of complexity) and correlate these to fault attributes (i.e. fraction of faults detected / isolated, fraction of false alarms and rate of false alarms) to develop a fault predictor using a Bayesian approach. The algorithm was used at the system design phase, to verify a proposed modification that is robust to fault throughout the operational life of the aircraft. 2) Neural network based intermitted fault detection: The reduced size of electronic chips and increases in their interactions and complexities have led to difficulties in diagnosing system faults; especially when these faults occur at the component level and are intermittent in nature. I developed a neural network that can be used and to successfully diagnose intermittent faults and provide intelligent reasoning on intermittent fault progression. The project was developed in-situ health monitoring technology to detect and characterise intermittent failures arising from dry solder joints, loose connections, damaged wiring and environmental degradation. 3) Estimating the impact of faults from a finance and safety point of view: This work was a continuation of the above two research results where the results were used to estimate the financial costs when faults occur. This was particularly difficult to obtain with very little information in the public domain - which means that the result was highly uncertain and hence further required a probabilistic approach to estimate its impact. Bayesian methods were used to factor in uncertainty. 4) After achieving the previous three goals I moved my focus on research to be implemented on a UAV fuel rig to demonstrate it practical significance. The main aim was predict the faults and its impact on a real application. I used a fuel rig was (using a proprietary diagnostic software called DSI STAGE Tool) to help understand the existing design limitations associated with system testability. A Failure Mode and Effects Analysis (FMEA) was defined to identify the critical (mission) failures. Using this information, different sensor placements were made to improve the fuel rig design. Finally, the impact of design changes was documented with regards to the overall reliability of the design.

I recently published a book about my work: http://books.sae.org/r-441/

I also published 7 journal articles in high impact journals! How cool is that… but it does show the importance of my research and my ability to contributed to scientific knowledge :p

For more information see: Buy my book!