Dynamics and Control

The dynamics and control research area carries out research across a wide range of dynamic systems from turbine blades through to aircraft landing gear.

Research aims

The dynamics and control research focuses on modeling and computing often non-linear systems to enhance their performance and support control system development. Many of our projects aim to use this knowledge to improve efficiency and work towards achieving Net Zero.

Research areas

Find out more about our research areas:

Dynamic analysis of structures with contact interfaces and mistuning effects

Most practical structures have contact interfaces and joints, leading to complex interactions in flexible rotor systems, especially with bearings or rotor-stator contact. These interactions can significantly alter dynamic responses, potentially causing multiple solutions, bifurcations, instability, or chaos.

In this research, models are developed to predict the dynamics of strongly nonlinear structures with friction, gaps, and contact interfaces.

Faculty

Yevgen Petrov

Renewable-powered hydrogen system control & optimisation

The current energy landscape reflects a complex interplay of challenges and opportunities. The awareness of climate change, global warming, and the depletion of fossil fuels has been the driving force for the transition to cleaner energy sources. Recent innovations, such as AI-powered systems, are further intensifying the demand, as they rely heavily on substantial and consistent energy inputs to enable the operation of data centres, machine learning models, and real-time analytics. It is not only generating more energy but also ensuring it comes from renewable and sustainable sources. In renewable energy systems, hydrogen has been shown to offer distinct advantages as an energy storage medium when compared to conventional energy storage mechanisms, such as battery technologies.

The research mainly focuses on the system integration, control, and optimisation of renewable power generation, especially using hydrogen as an energy vector for storage. This includes the control and optimisation of individual components such as photovoltaic panels, wind turbines, electrolysers, batteries, and fuel cells, as well as the integration of those components to form an energy generation, storage and utilisation system, determining the configuration, the optimal size, and developing an overall energy management strategy to coordinate the operation of those components to meet the requirements defined by specific applications, e.g. efficiency, penetration of renewable power sources, overall techno-economic benefits etc. The research also looks into wider extension of this renewable-powered hydrogen system, such as power-to-gas technology, vehicle-to-grid applications, as well as how AI technologies can further enhance its performance.

Faculty

Fan Zhang

Uncertain structural dynamics applied to extreme vehicle motions

Predicting the dynamic response of structures with uncertain parameters is crucial for safety and design reliability, especially in vehicle dynamics. Catastrophic events like vehicle lift-off and flip-over, caused by unfavorable suspension parameters, are rare but critical to predict. Vehicle dynamic models are essential for manufacturers since testing every scenario isn't feasible. However, predicting these rare events typically requires lengthy Monte Carlo simulations, which are prone to statistical noise.

This research is developing a precise and efficient method to predict rare vehicle flip-over events without statistical noise, avoiding the heavy computational demands of Monte Carlo simulations, even with many uncertain suspension parameters.

Faculty

Julian Dunne

Optimal control of autonomous vehicles

Autonomous vehicles (AVs) are being developed to reduce traffic accidents and expand mobility options, but they are highly energy-intensive due to the many on-board sensors. Efforts are being made to reduce this energy consumption without sacrificing safety and comfort. One approach is optimizing steering control, though the nonlinear nature of steering models complicates this.

Ongoing research aims to develop robust optimal control methods to effectively reduce energy use in AVs.

The perception module plays a crucial role in the autonomous driving system, which is a very complicated system. In contrast, current perception research in the area of autonomous driving primarily focuses on the recognition of vehicles, lanes, and traffic signs, ignoring other factors that may cause traffic accidents, and fails to consider the fact that many traffic accidents on highways are caused by wild or wandering animals. The following studies were conducted to close this gap in knowledge:

  • a dataset of 1050 photos for large animals that could appear on highways
  • a more efficient Yolo model by improving its backbone, replacing the C3 module with C3Ghost. The number of parameters is decreased to fewer than 3.7 million, just 52.7% of Yolov5s but the average accuracy for each type of animals (mAP% 0.5) has reached over 95%
  • our GhostSort-YoloNet (GS-YoloNet) also incorporates the Deep Sort algorithm to achieve real-time ranging and speed assessment of numerous targets, which is promising for practical application.

Faculty

Julian Dunne
Bao Kha Nguyen

Autonomous driving with end-to-end deep learning

Autonomous driving systems commonly adopt a modular approach, dividing the system into different modules with specific functions to address various tasks.  However, the modular execution approach means that any errors in one module can propagate and affect subsequent modules, ultimately impacting the performance of the entire system.  The end-to-end approach has gained widespread attention owing to its ability to simplify the design of autonomous driving systems by directly learning from sensor data to vehicle control commands (such as acceleration, steering, braking, etc.).  This research aims to improve the existing end-to-end approach, with a particular focus on occupancy grids and maps in perception.  It aims to directly generate occupancy maps from sensor data and build an end-to-end perception model architecture to predict and reconstruct surrounding scenes.  This will optimise the prediction, planning, and decision-making capabilities of the existing published end-to-end systems, thereby improving the overall performance and safety of autonomous driving systems.

Faculty

Fan Zhang

Power from hydrogen-fuelled rotary resonating free piston expanders

Free-piston engine generators offer several important benefits over conventional generators comprising an IC engine directly coupled to an electrical generator. These benefits include higher efficiencies, greater compactness, and lighter weight, resulting in significantly higher gravimetric and volumetric energy densities, compared to alternatives. In addition, the important possibility of electrically-controlled variable compression ratio, offers fuel flexibility, including the potential to run on zero carbon fuels such as hydrogen. Resonating opposed-piston FPEGs offer significant advantages over FPEGs fitted with a bounce chamber. These advantages include total dynamic balance, greater linearity allowing more precise control, no energy loss through bounce chamber compression, and a significant reduction in electrical machine current, with a corresponding reduction in losses. The fitting of a stiff resilient member to an FPEG, creates a mass-elastic system capable of mechanical resonance.  Resonating rotary opposed piston FPEGs also have the advantage of not suffering from asymmetry, and better potential to use labyrinth sealing instead of a ring-pack to avoid unburned hydrocarbons from lubricants (plus their polluting additives). This research is exploring the full potential of hydrogen-fuelled resonating rotary free-piston generators in addressing hard-to-decarbonize transport applications. 

Faculty

Julian Dunne
Cyril Crua
Spyros Skarvelis-Kazakos
Mark Puttock-Brown