2024 Capability Portfolio Digital | Page 185

Clean Energy Capability Portfolio | Empowering Consumers

Reinforcement Learning

for Operations and
Maintenance Planning

World-leading Research Group in the field of machine condition monitoring that brings together two primary areas of expertise – wear debris and vibration analyses – for applications in the field of machine condition monitoring .

Competitive Advantage

• Ability to integrate wear and vibration analysis to monitor machine condition , and predict the remaining life of critical assets
• Cutting-edge diagnostic and prognostic tools to inform maintenance decision makers
• Expertise across a range of technologies , including signal processing , wear analysis , and digital twin and Artificial Intelligence ( AI ) -based techniques
• Collaboration with universities and industries on a domestic and global level

Impact

Advanced techniques for generating significant economic benefits and increased safety for personnel by :
• Providing an early warning of possible faults and / or failures of mechanical components ( e . g ., gears and bearings )

Capabilities and Facilities

• One-stage and multi-stage gearboxes , and bearing test rig with multiple sensors , to generate data under various health and operating conditions
• Wear testing and analysis facilities and expertise
• Advanced vibration signal processing techniques , including torsional vibration and transmission error , for fault detection and diagnosis of critical components ( e . g ., gearboxes , bearings , IC engines )
• Integrating tribological and vibration information to achieve continuously updated wear assessment and prediction models , for condition monitoring and remaining useful life ( RUL ) prediction
• Using continuously updated digital twin models , and AI technology , to simulate machines and their components in healthy condition , and with faults of different types , severity , and locations
• Using reinforcement learning for operation and maintenance schedule optimisation
• Diagnosing their severity
• Predicting their remaining useful life
• Optimising their operation and maintenance using deep learning approaches , such as reinforcement learning

Our Collaborators

• Defence Science and Technology Group ( DSTG )

More Information

Professor Zhongxiao Peng

Successful Applications

School of Mechanical Engineering
• Gearboxes used in defence forces and aerospace
T : + 612 9385 4142
• Drivetrains in wind turbines
E : z . peng @ unsw . edu . au
• Gearboxes , bearings , engines used in transportation , mining processes , and agriculture

185