In earthquake-prone regions, monitoring for faint tremors can predict an upcoming disaster. In a similar way, monitoring electrical transformers or bridges for the ‘tremors’ of accelerated degradation can significantly extend their working life in a safe way, says Prof Tshilidzi Marwala from the University of Johannesburg (UJ).
On 14 October 2016, The Alumni Association of Case Western University in Ohio, USA honoured him with its 2016 Professional Achievement award for achieving as the Deputy Vice-Chancellor: Research, Postgraduate Studies and the Library at the University of Johannesburg; being well-regarded in South African government and industry realms; and leading delegations to the South African Parliament. Case Western is associated with 16 Nobel prize-winners. Currently the university places 126th globally in the Times Higher Education rankings.
Prof Marwala and his research team are developing innovative ways of remotely monitoring critical national infrastructure such as transformers and bridges in real time. They feed this monitoring data into real-time Artificial Intelligence (AI) machines to detect faint indications for impending failure and to recommend predictive maintenance before breakdowns can occur.
“If a transformer fails, an entire area such as Sandton, which is one of Gauteng’s major city centres, can stop functioning. Generally, power failures happen more frequently because of a transformer failing rather than a power station not providing electricity. Transformers are expensive, costing R30-million upwards. On top of that, the manufacturers recommend replacing them after three to five years,” he says.
“However, you can remotely monitor an oil-filled transformer for temperature fluctuations and gas emissions in real time, 24 hours a day, over a period of two or three years. Humans aren’t good at interpreting all this data though. A person that has to watch the data from one sensor in one dimension can still cope. Give the same person 10 sensors to watch and they’re no longer functional. When you have 20 sensors on a transformer or a bridge, monitoring all that multi-dimensional data has to be done by a computer,” he adds.
When a power provider sticks with a simple maintenance schedule, it ends up removing a transformer that is usually still working well, and replacing it with a new one. That old transformer is over-designed, meaning it can still operate for longer, but needs to be monitored to see if its wear and tear is accelerating, which means it can fail soon. Picking up the faint signals of accelerating degradation is what Artificial Intelligence machines are particularly good at.
“Even if a sensor fails, and does not send through its data, the AI machines we have developed can still make accurate predictions with the incomplete data set,” says Prof Marwala.
“By monitoring the structural integrity of the transformer in this way, it can be operated safely and predictably well beyond the original scheduled maintenance removal date. In effect, the transformer is on a predictive maintenance schedule, where the predictions are provided by the sensors and AI machine,” he says. German manufacturer Siemens has commended Prof Marwala’s work in this area as a leading project globally.
Research interests, recent awards, academic qualifications
Prof Marwala’s research interests include the application of computational intelligence to engineering, computer science, finance, social science and medicine. His work has been featured in magazines such Time Magazine, New Scientist and ACM Tech News.
He published on using Artificial Intelligence in military conflict in the book Militarized Conflict Modeling Using Computational Intelligence in 2013 with Springer London.
Prof Marwala has authored or co-authored nine books. His research with PhD and Master’s students at UJ is ongoing. He has supervised, to completion, 20 PhD and 47 Master’s theses.
Prof Marwala matriculated at Mbilwi Secondary School in Limpopo in 1989.
He holds a Bachelor of Science in Mechanical Engineering (Magna Cum Laude) from Case Western Reserve University in the United States; a Master of Engineering from the University of Pretoria; a PhD in Engineering from St John’s College, Cambridge UK; and successfully completed a Program for Leadership Development at Harvard Business School, USA. In 2006 to 2007 he was a visiting fellow at Harvard University. In the year 2007 to 2008, he was a visiting fellow of Wolfson College, Cambridge.
