PhD Students

Railway bridges operate under constant loads and varying environmental conditions, which can gradually affect their structural performance. Early detection of these changes is important to ensure safe and reliable operation. Structural Health Monitoring (SHM) systems offer continuous data that can support this process, although turning large volumes of raw measurements into useful information for maintenance is still challenging.

My research focuses on developing a practical approach to analyse SHM data from a steel truss railway bridge in Indonesia. The bridge is monitored by 28 sensors that record displacement, strain, acceleration, and temperature. These field measurements are combined with a Finite Element Model (FEM) to better understand the bridge’s dynamic behaviour and to identify signs of deterioration or unusual loading.

By applying risk-based methods, the project aims to evaluate deterioration patterns, reduce uncertainties, and support more informed maintenance decisions. The broader goal is to contribute to a more efficient and reliable asset management process for railway bridges.

Before starting my PhD, I worked as a Transport Infrastructure Engineer and researcher at the National Research and Innovation Agency (BRIN), where I was involved in structural monitoring and infrastructure management activities.

Signalling is one of the most significant parts of railway safety and operational efficiency, ensuring that trains move safely and smoothly across the network. Telecoms support real-time communication between trains, maintenance teams and control centers. The performance of telecoms asset involves not only the equipment's physical reliability, but also multitude of factors, including life cycle, maintainability, and workforce competency levels, all of which can lead to obsolescence. Obsolescence closely related to spares management policy, due to outdating of asset components that can cause downtimes, unreasonable maintenance and at last ineffective operation of the system.

During this PhD project, my target is to create an obsolescence modelling framework for asset management of railway signalling and telecoms systems, that includes spares planning, considering regional differences and workforce. In addition, I am developing a vision to specify a multi-objective model structure aimed at balancing whole-life cost, safety, performance, sustainability and maintenance.

Before I started the PhD research journey, I graduated Master's diploma as an engineer of railway stock at Dnipropetrovsk National University of Railway Transport in Ukraine. Solid background of railway rolling stock helps me to understand and proceed with current project. Furthermore, I had experience working as Engineering Specialist at JSC Ukrainian Railway, that developed my skills in data analyzing and making different types of reports.

Asset management and maintenance of rail track assets often rely on data collected using remote condition monitoring approaches. Through wear due to freight or passenger trains passing over the track, different faults can develop on the track that can lead to serious safety and performance risks. Railway infrastructure providers work to predict these faults by utilising monitoring methods that collect different modes of data, including digital imagery, track geometry data, and other types of data depending on the scenario.

Inspired by developments in computer vision, this research examines the potential viability of incorporating multimodal neural networks for the prediction of track condition using high-definition photographs and historical track geometry readings among other modes of data. This approach aims to incorporate the relevant contextual data with historical readings into the prediction of the future track condition. This work intends to aid to regional engineers to help monitor track assets and thus make decisions on what maintenance is most needed.

Before starting the PhD in October 2024, I completed a integrated masters of mathematics at the University of Nottingham. The focus of my masters dissertation was using Bayesian Statistics for epidemic and population modelling. Areas of interest are machine learning, computer vision and statistics. 

It is obviously desirable to maximise performance of the Hydrogen Refuelling Stations not just for economic reasons, but to deliver the best customer experience. This project seeks to optimise the plant operation, where planning preventative maintenance can help reduce disruption to service and improve the commercial case of a plant.

Salim graduated from The University of Huddersfield where he studied Electrical/Electronic Engineering. He completed his MSc in Powertrain Engineering from IFP School (École Nationale Supérieure du Pétrole et des Moteurs) in 2020. His interest is on hydrogen technology infrastructure and vehicle applications.

High-speed railway infrastructure is a complex arrangement of systems and structures, which includes: track, switches, drainage, signalling, power supply and communications, in addition to the civil structures comprising earthworks, tunnels, bridges and stations. As new high-speed railways are built, it is important that plans are in place to ensure they are sustainable and affordable so that elements can be renewed as they wear out or become obsolete due to new technologies. The funding for renewals is generated through a usage charge levied by the infrastructure owners on the train operating companies. However, should this charge be too little then necessary replacements will not take place, or, in the event that it be too much, the consequence is higher than necessary fares for passengers.

To be able to fix and justify the right charge requires advances in areas of engineering. It is necessary to understand how all the infrastructure elements degrade due to either the passage of time or use. This enables the estimation of when replacement is necessary, and can be achieved using modelling methods which include artificial intelligence (AI). For systems and structures made of many components, aging at different rates, there is the additional challenge of combining the component performance predictions to give the performance of the entity.

Having established the degradation mechanisms for each of the asset types it is then necessary to plan when the renewals will be performed. This decision will be based on the costs incurred when selecting the alternative decisions and can be optimised. Some assets will degrade more slowly than others and so have more flexibility in fixing the exact renewal date. There are also practical considerations which will minimise any consequential service disruption.

Utilising the asset degradation models enables the asset renewal schedule to be produced in such a way as to minimise the costs to the infrastructure owner. This will be achieved through the definition of an optimisation problem which minimises the whole system costs and satisfies constraints which account for the practicalities of performing the renewals to have minimal impact on the service provision. The costs included in the objective function will include the costs of the renewals, the costs of the penalties for service disruption and the costs associated with additional maintenance to keep aging assets in an acceptable condition beyond the predicted renewal time. The modelling will account for the uncertainty in both the costs and the degradation profiles. Once the model is created it would enable the new schedules to be determined in the event that track utilisation (volumes of freight and passenger traffic) changed. 

The UK's Rail Technical Strategy (RTS) 2020 sets a 20-year vision for railway service quality with a focus on safety, reliability, and resilience, as well as meeting capacity and service requirements through innovation and technology. This is a major challenge when, in addition to component or system failures and human errors, modern railways also experience threats such as cyber-attacks, natural hazards and climate change. Current rail industry practice for choosing recovery actions after a disruption is based on pre-written contingency plans and, to a degree, relies on controller judgement, but the various options cannot be explored automatically. To ensure quick, effective recovery and minimise disruption, there is a need for an optimization-based decision-making procedure to be developed.

This work presents a probabilistic mathematical model to analyse and predict water distribution network resilience, aiding managers in optimal repair decisions during disruptions. The model evaluates the network's ability to withstand and recover from disruptions, considering parameters like pipe diameter and probabilistic factors such as demand fluctuations, failure rates, and mean time to respond. The authors underscore that integrating a probabilistic model, considering real-world uncertainties, is crucial for optimizing decision-making in management and repairs and enhancing water network resilience.

I am currently conducting research aimed at improving the locomotive fleet management system. The locomotive/train fleet management system is complex, requiring innovative approaches and methodologies to maintain the high productivity of the system. In the research, a model of the system will be developed to predict the effectiveness of different train maintenance strategies. The model will be applied to the fleet of trains used to provide service on a railway network. The prediction will account for practical limitations in the financial and equipment resources of the system aiming to optimise service provision.

The condition of bridges and their components is the result of two key processes: deterioration and maintenance. To effectively maintain any network, asset managers typically use life-cycle models to forecast the condition of bridges and bridge components, enabling effective allocation of resources to maximise condition and safety and minimise risk. Accurate modelling requires detailed understanding of both processes. However, data on interventions and their impact is often sparse and incomplete, leading to inaccurate estimations of the impact of maintenance work. This makes it difficult to separate the contribution of the two processes to the condition record, leaving artefacts of undetected interventions and causing slower predictions for deterioration rates.  

In this research, a novel approach is being undertaken to model the behaviour of defects on bridge components through progression of deterioration; inspection; decisions on the application of maintenance work; and repair and replacement activities. Using Coloured Petri Nets to model defects on a portfolio of bridges, the model simulates real-world behaviour and practices taken by Network Rail, who own and manage over 26 000 railway bridges in the UK.   

This work uses extensive real condition data and limited intervention records to calibrate deterioration rates, and draws upon detailed industry expert knowledge to replicate the processes and logic taken by Network Rail in applying maintenance work, to create a comprehensive forecasting model. The project aims to output deterioration rates with improved accuracy than the current approach, and a detailed understanding of intervention processes and effectiveness

  Improving reliability and preventing failures are not only applied to manufactured products, but also to intangible objects such as medical processes. Prioritising patient safety is covered by the reliability of the medical processes. However, a process or procedure cannot be separated from deviations, which influence process reliability and increase the safety risk to the people involved in the procedures. Deviations may emerge from various sources, such as human errors, patient condition, hospital quality, and many more. Hence, identifying deviations manually is challenging due to data limitations, and numerous causes of deviations need to be explored. 

  This research proposes a novel method that integrates system modelling and artificial intelligence (AI) techniques to efficiently map and capture variations in medical processes to determine the outcome for the patients. The proposed model will be focused on the Newborn Life Support (NLS) procedures and possibly be extended to other types of medical procedures. As the system modelling approach, a Coloured Petri Net (CPN) will be utilised to mimic the actual NLS procedures. Numerous aspects involved in the procedures will also be considered using deterministic and/or probabilistic rules to create a realistic model. Furthermore, the trained machine learning and/or deep learning models will take part in ruling out the mechanism of the simulation by automatically identifying deviations in the procedures. Finally, this research will contribute to providing a deeper insight into how the variations in clinical procedures behave and affect the outcomes, which is the main parameter of its reliability. The utilisation of AI models will also provide reproducibility and accuracy for practical application in the future. 

 Brief background description: 

Fakhri obtained his master's degree in the Department of Industrial Engineering and Management, Bandung Institute of Technology, Indonesia. His research areas are reliability engineering, quality engineering, and artificial intelligence applications. Highlights of his previous research projects included a funded research project in developing a failure interaction model for automobile products that has been applied to optimise the warranty cost under the Lemon Law.  Another research project that he has accomplished is developing a predictive maintenance policy using a machine learning architecture for industrial machinery. Finally, his current research interest is (but not limited to) reliability engineering of a system by applying artificial intelligence models, which can be applied to tangible (e.g., manufactured products) or intangible (e.g., service industry) systems.