Intelligent tool makes aviation engine maintenance easier
Engineering and maintenance associated with aviation engines may be much easier and cost-effective in the future, with the Queensland government awarding a Brisbane-based company $48,527 in funding to build a machine-intelligent modelling tool.
The average aviation engine costs around US$16.25 million each, and any failure and maintenance can be very expensive, as it means a grounded plane and lost revenue, or worse, an accident.
The high costs associated with engine failure or damage are compounded by the complexity of the turbine engine, especially when it is in operation. Small changes in engine thermodynamics and physical dynamics can have a cascading effect.
Evolving Machine Intelligence (EMI) aims to use advanced algorithms and noise-polluted partial information to build up detailed mathematical models of these complex dynamic systems, so it is possible to understand what is happening without stopping or dismantling the engine, and determine if a unit requires intervention, and if so, the nature of that intervention.
Towards that end, EMI has won funding from the Advance Queensland Knowledge Transfer Partnerships program, which is run by the Queensland State Government. The grants program aims to enable businesses to partner with universities to engage graduates on innovative projects.
EMI describes itself as a pure research and development entity which is developing machine intelligence technologies that will allow machines to understand and control complex systems, including other machines, non-invasively.
EMI founder Dr Nigel Greenwood has been working in machine intelligence for two decades, developing technology that allow complex systems to be reconstructed from partial data. A top aviation engine manufacturer has already funded a demonstration project of his technology.
According to the organisation, it is using a new technique called “phi-Textured Evolutionary Algorithms” to reconstruct and evolve candidate mathematical models from the available information about complex systems.
The models enable the testing of assumptions and ambiguities, exploring different avenues and possibilities, generating strategies for system control, communicating this information to human operators, and implementing the control strategies.
The proprietary algorithms are powered by new massively-parallel GPU cards and other high-performance computing hardware.
This technique is said to be superior to neural networks and fuzzy logic (other major contenders in machine intelligence research and development).
While the technology has potential applications in oil and gas, power generators, and industrial systems, EMI has a commercial spin-off company called Turbine MachineGenes (TMG) which focuses on machine-intelligent reconstruction of engines and other industrial systems. Investors in TMG include UniQuest, the main commercialisation company of The University of Queensland.