Page 47 - University of Pretoria RESEARCH REVIEW 2018
P. 47

Reshaping optimality Daniel Nico Wilke, Department of Mechanical and Aeronautical Engineering
Finding optimal process and design engineering solutions requires models that are computationally efficient, and exploit the right information.
 Increasingly, in global and local economic contexts where the 4th Industrial Revolution (4IR), artificial intelligence and the ‘internet of
things’ have become pervasive,
the efficient management of large production volumes necessitates precise engineering solutions. This dynamic and fast-changing context places significant pressure on design and process engineers to find the best solutions, to propose competitive approaches to production and service, and to deliver new products timeously. To achieve these increased demands, engineers need responsive and accurate computational models.
In real-world business and engineering contexts, time is of the essence. Weeks, even months, are often required to obtain an approximate answer to assess the quality of a single design or process. Enabling design and process engineers to formulate questions accurately, and to obtain answers within short turnaround timeframes, used to be unthinkable where dilatory systems are pervasive.
Nico Wilke, Associate Professor
in the Department of Mechanical and Aeronautical Engineering, has worked on enhancing the efficiency of design and process engineers
by developing a mathematical framework that can handle estimates of vastly different accuracies and consequently computational costs. He has also worked with researchers in moving, with ease, from smooth and continuous classical optimisation, to difficult and discontinuous functions, by using gradient-only optimisation and gradient-only surrogates. This work has had application in contexts ranging from finding optimal shapes of structures, to automating the training of deep learning models on graphical processing units.
Wilke’s co-authored book, Practical Mathematical Optimization (2018, Springer, 2nd edition), was the first on the international market to cover this particular mathematical framework and computational model, supported by unique optimisation algorithms. The focus on application is intended to save valuable computational time for engineers and researchers in solving specific engineering problems. These ideas have also been presented on the international stage through invited talks and PhD courses.
In a number of international collabo- rations, significant progress has also been made to simulate processes that involve granular materials. Although
the second most manipulated substance after water, they remain one of the least modelled by engineers. Significant contributions have included: i) accelerating simulations, ii) improving particle shape representations, and iii) interactive simulations, being enabled by computing on graphical processing units. Over the last three years, commercial codes have followed suit in the use of graphical processing units.
 Nico Wilke received the 2018 Vice-Chancellor’s book award for his book on mathematical optimisation, co-authored with the late Jan Snyman.
45
 




















































































   45   46   47   48   49