The aim of the research was to improve the fundamental understanding of mechanical behaviour of ballast and study the effect of tamping on ballast. The experiments were carried out on scaled railway ballast to eliminate the difficulties associated with testing large particle granular materials. Consideration was given to the gradation, mineralogy and shape during scaling. Particle characterisation work was carried out on scaled and full size ballast using imaging techniques to examine the validity of the use of scaled ballast. Detailed analysis of results is used to quantitatively measure the changes in shape with particle size. The results show measurable differences in particle shape between different particle size intervals. As the differences are small in magnitude, they do not invalidate the use of scaled ballast. Monotonic, cyclic experiments were carried out as part of a laboratory testing programme. Scaled ballast shows generally similar stress strain behaviour to larger granular materials. The friction angle of 40o to 50o for the confining pressure range of 15 kPa to 200 kPa falls within the range of friction angle obtained for full size ballast in the literature. The effect of changing confining pressure during train loading was examined. The results show that the settlement increases and the stiffness reduces when the confining pressure cycles. The effect of principal stress reversal during tamping was examined by an extension stage after the cyclic loading. It is shown that massive settlement occurs after the extension stage during initial cycles and settlement returns back to the pre-extension stage soon under loading. The results evidence the disruption of ballast structure and loss of stiffness due to tamping. Specimens were resin stabilised within the triaxial cell after specific stress paths had been followed. The changes in structure during a specific stress path can be studied by CT examination.

For high-speed trains, aerodynamic noise becomes significant when their speeds exceed 300 km/h and can become predominant at higher speeds. Since the environmental requirements for railway operations will become tighter in the future, it is necessary to understand the aerodynamic noise generation and radiation mechanism from high-speed trains by studying the flow-induced noise characteristics to reduce such environmental impacts. The aim of this thesis is to investigate the flow behaviour and the corresponding aeroacoustic mechanisms from high-speed trains, especially around the bogie regions. Since the prediction of the flow-induced noise in an industrial context is difficult to achieve, this study focuses on scale models with increasing complexity. The aerodynamic and aeroacoustic behaviour of the flow past an isolated wheelset, two tandem wheelsets, a simplified bogie and the bogie inside the cavity with and without the fairing as well as considering the influence of the ground are investigated at a scale 1:10. A two-stage hybrid method is used consisting of computational fluid dynamics and acoustic analogy. The near-field unsteady flow is obtained by solving the Navier-Stokes equations numerically through the delayed detached-eddy simulation and the source data are applied to predict the far-field noise signals using the Ffowcs Williams-Hawkings acoustic analogy. All simulations were run with fully structured meshes generated according to the guidelines based on a grid independence study on a circular cylinder case. Far-field noise radiated from the scale models was measured in an open-jet anechoic wind tunnel. Good agreement is achieved between numerical simulations and experimental measurements for the dominant frequency of tonal noise and the shape of the spectra. Numerical results show that turbulent flow past the isolated wheelset is dominated by three-dimensional vortices. Vortex shedding around the axle is the main reason for the tonal noise generation with the dominant peak related to the vortex shedding frequency. The noise directivity shows a typical dipole pattern. Moreover, for both the tandem-wheelset and the simplified bogie cases, the unsteady flow developed around them is characterized by the turbulent eddies with various scales and orientations including the coherently alternating shedding vortices generated from the upstream axles. The vortices formed from the upstream geometries are convected downstream and impinge on the downstream bodies, generating a turbulent wake behind the objects. Vortex shedding and flow separation as well as interaction around the bodies are the key factors for the aerodynamic noise generation. The radiated tonal noise corresponds to the dominant frequencies of the oscillating lift and drag forces from the geometries. The directivity exhibits a distinct dipole shape for the noise radiated from the upstream wheelset whereas the noise directivity pattern from the downstream wheelset is multi-directional. Compared to the wheelsets, the noise contribution from the bogie frame is relatively small. Furthermore, when the bogie is located inside the bogie cavity, the shear layer developed from the cavity leading edge has a strong interaction with the flow separated from the upstream bogie and cavity walls. Thus a highly irregular and unsteady flow is generated inside the bogie cavity due to the considerably strong flow impingement and interaction occurring there. Unlike the isolated bogie case, noise spectra from the bogie inside the cavity are broadband and a lateral dipole pattern of noise radiation is generated. The noise prediction based on the permeable surface source is formulated and programmed using the convective Ffowcs Williams-Hawkings method. Results show that the bogie fairing is effective in reducing the noise levels in most of the frequency range by mounting a fairing in the bogie area; and for the bogie inside the bogie cavity with the ground underneath, the far-field noise level is increased due to more flow interactions around the geometries and the ground reflection effect.

Despite compacted soils being unsaturated, geotechnical design is predominantly based on saturated soils, despite it being known that differences exist between the behaviour of saturated and unsaturated soils. Suction in soils has been shown to increase its strength and stiffness with the increase dependent on the degree of suction, soil type and particle arrangement. Future climate change predictions suggest an increase in average global temperatures leading to longer drier summers and wetter winters resulting in seasonal fluctuations in suction, such that unsaturated soil behaviour will become more important. Transport infrastructure such as railway tracks, founded at shallow depths, are likely to be unsaturated and influenced by variations in suction. A key parameter in assessing the performance of soils is stiffness; in particular shear modulus, however, although suction affects shear modulus, limited research has been conducted into its influence. This research, therefore, considers how variations in suction influence the small strain shear modulus of railway formation material. Soil water characteristic curves were determined, using the pressure plate apparatus and filter paper technique, for four materials covering the spectrum of railway formation materials encountered on the COALlink line in South Africa. Shear modulus measurements of specimens prepared from two of these materials were determined using a Stokoe resonant column apparatus. Suction was controlled by preparing specimens at different water contents, with the suction independently measured using the filter paper technique. Results showed that shear modulus was significantly influenced by suction and exhibited a complex behaviour. For specimens tested at their preparation water content shear modulus increased with increasing suction up to an optimum value, and then reduced as suctions exceeded this optimum value forming a bell-shaped curve. The influence of suction was seen to be greater at lower net normal stresses. Although this behavioural pattern was similar for both materials tested in the resonant column apparatus, the peak value of shear modulus increased with clay content. Importantly, the marked changes in shear modulus due to suction coincided with the suction range predicted in-situ. Tests were also conducted on specimens prepared at different water contents but the same density, airdried to very low water contents. The shear moduli of these air dried specimens were markedly higher than those of the dry and unsaturated specimens tested at their preparation water contents. Variation in measured shear modulus between specimens led to the conclusion that different particle arrangements between the specimens may have also contributed to the changes in shear modulus and were dependent on the preparation water content. At high water contents the clay particles became evenly distributed around the sand particles when the specimen was prepared, whilst, at low water contents the clay formed lumps during preparation of the material mix and became embedded with the sand particles during formation. CT scanning undertaken to look at the particle arrangement highlighted regions throughout the specimen where sand particles were embedded in lumps of clay and silt particles. These results are the first to demonstrate the importance of suction and its variation on the small strain shear modulus behaviour of railway formation material.

This thesis adopts a holistic approach towards railway track capacity to develop methodologies for different aspects of defining, measuring, analysing, improving and controlling track capacity utilisation. Chapter 1 presents an overview of the concept of capacity and the railway capacity challenge is explained. Chapter 2 focuses on past approaches to defining and analysing the concept of railway capacity. Existing methods for estimating capacity utilisation are studied in four categories: analytical methods, parametric models, optimisation and simulation. Chapter 3 examines various factors affecting capacity utilisation. Chapter 4 develops the systems engineering foundation toward railway capacity. From process improvement methods, Six Sigma and its Define, Measure, Analyse, Improve and Control (DMAIC) cycle is chosen as the underlying framework of the thesis. Chapter 5 defines lean, micro and macro capacity utilisation based on the discrete nature of railway capacity. Data Envelopment Analysis (DEA) is used to develop two novel methodologies to analyse lean capacity utilisation. A DEA model analyses relative efficiency of train operating companies based on their efficiency to transform allocated train paths (timetabled train kilometres) and franchise payments to passenger-kilometres while avoiding delays. A case study demonstrates its application to 16 train operating companies in the UK. The operational efficiency of stations is benchmarked from similar studies for ports and airports. Two models are developed for analysing technical efficiency and service effectiveness. 96 busiest stations in Great Britain are analysed by this method. For analysing capacity utilisation in the freight sector, the concept of ‘profit-generating capacity’ is introduced in chapter 6. It is applied in an American freight case study to choose between bulk and intermodal trains in a heterogeneous traffic. DEA is also used in another case study for identifying the most profitable commodities. Chapter 7 suggests using variation reduction and failure mode and effect analysis (FMEA) to control capacity utilisation. For improving railway capacity utilisation it is suggested to find and improve the weakest line section, the weakest trains and the weakest station. A real world case study of the South West Main Line in Great Britain, demonstrates applying these aspects. For finding the weakest line section two existing methods of the UIC 406 and the CUI method are compared with each other. For finding the weakest trains a meso index is suggested. It can identify which trains can be removed to free up some capacity in the busiest section of the line. Simulating delays and removing the highest delay causing trains is another method suggested. The weakest stations are identified by applying the DEA methodology developed in chapter 5. Target values for train stops at each station are suggested to be fed to the tactical timetabling. It is concluded that developing methodologies to analyse, improve and control railway capacity utilisation is needed and the methodologies proposed in this thesis can be a stepping stone towards them.

Ballast is a primary component of the railway track and it has several functions including the ability to resist vertical, lateral and longitudinal forces applied to the sleepers from the train wheels. In response to the stresses induced during a train passage, ballast typically experiences plastic settlement which can reach unacceptable magnitudes. Thus, ballast is the focus of the majority of maintenance and renewal activities. In response to the growing need for resilient track materials (including ballast) to cope with increasing train speed, load and frequency, means of optimizing its performance and minimizing maintenance requirements are required. Ballast improvement techniques, such as the use of geogrids, have attracted a great deal of attention in terms of research, but huge potential for reducing costs through ballast improvement still exist. It has been shown from previous research that the use of randomly oriented fibres in sands can significantly improve their strength. This technique might be used to improve ballast strength and reduce track geometry deterioration. In addition, fibre reinforcements in ballast can potentially provide a new method of reinforcing ballast whilst maintaining the capability of withstanding typical tamping operations which are incompatible with geogrids. However, there is a lack of rigorous scientific understanding of the effects of fibre reinforcements on relatively large aggregates such as railway ballast. This research examines the effect of random fibre inclusions on the packing structure of granular materials (coarse sand and scaled ballast) and describes the physical implications associated with the observed changes. The mechanical properties of fibre reinforced granular materials across different particle sizes and a hypothesis of fibre/particle reinforcement suitable for relatively large particles are presented. An image-based deformation measurement technique has also been used to investigate the effects of fibre reinforcements on the local deformation of triaxial specimens to corroborate the observed mechanical behaviour of the reinforced specimens. Finally, the mechanical behaviour of a fibre reinforced ballast layer below a cyclically loaded railway sleeper was investigated using a full scale laboratory tests.

The European Train Control System (ETCS) as part of the ERTMS (European Rail Traffic Management System) is a train control and automation system, which has been introduced into the UK rail network. The major change with the introduction of the ERTMS has been the shift of the movement authority from signals outside on the tracks to inside the cab, and the introduction of speed profiles that the drivers must adhere to. The introduction of this new system triggered the need to understand its impact on the train driving task and train driver behaviour. In particular, the effect the ERTMS has on drivers’ cognitive strategies and demands. The overall aim of the thesis was to understand the effects of new train control and automation technology on train drivers’ behaviours. The research was conducted, using a mixed methods approach, in the rail environment with train drivers and rail experts. Literature reviews of existing train driving models, train driving research and associated issues of increasing control and automation on human behaviour; were used to provide the theoretical context for the thesis. The literature review highlighted that there was a potential shift in cognitive strategies and demands with the introduction of train automation and control technology. Due to the limited amount of research in train driving on the whole, but also train driving with automation and control technology, the majority of the literature hypothesised the possible impact of the introduction of automation and control technology. An exploratory study of some of the different forms of train driving in the UK, was used to generate insight about train driving with different forms of train technologies and provided the theoretical foundations for the following studies. The emerging cognitive themes also addressed the gap in knowledge about train driving with different forms of technologies. The emerging cognitive themes from this study included route knowledge and memory, monitoring, allocation of attention, anticipation, prioritisation and decision making. A semi-structured interview study with ERTMS drivers, addressed some of the questions raised in the previous study using ERTMS drivers’ subjective experiences. Since the exploratory study, the results demonstrated an adaptation and shift towards acceptance of the system and it also identified some of the driving strategies that had emerged. This chapter investigated drivers’ subjective experience, highlighting high-level strategy changes. A real world exploratory eye-tracking study with both conventional and ERTMS drivers on their normal timetabled routes, provided a wealth of data. The first level of the quantitative eye-tracking analysis, aimed to address the industry question of ‘heads up’ vs. ‘heads up, heads down driving’. It demonstrated a shift of typical visual attentional strategy from monitoring outside on the tracks to speed information inside the cab. Analysis of verbal protocol data collected in the eye-tracking study also provided some rich qualitative data about train driver strategies and demands. Further analyses of the eye-tracking data, identified events where there is a difference in visual behaviour between ERTMS driving and conventional driving, but also between each type of driving and its own baseline data. An expert elicitation workshop with ERTMS human factors experts, was used to generate requirements for a future train driving model. The main findings highlighted that several models are needed to help address some of the issues raised, as they could provide different uses, acting as ‘building blocks’ to the overall picture. Qualitative models can be used to provide the framework and language as a communication tool, whilst more quantitative models can be used to compute error and workload. Models need to be informed by cognitive theory but also focus on the train driving tasks and information used by train drivers. Finally, the studies presented in this thesis were used to develop an integrated human factors model of the influence of train automation and control technology on train driving and guidance was generated for future train driving models for both conventional and ERTMS train driving. The current research has contributed critical knowledge to both the academic literature but also informed human factors practitioners in the rail industry. The thesis has contributed novel understanding about train driving with a control and automation technology, which have already been utilised by the rail industry.

In today’s modern world, society are accustomed to disposable products, temporary services and frequent replacements. The art of maintaining and renewing assets has been somewhat lost. However, in the pursuit of financial performance, comes the need to effectively manage assets. Management of a large portfolio of infrastructure assets is a complex and demanding task for infrastructure owners. Not only is the coordination of a large organisation difficult to align, but every decision is scrutinised by regulatory bodies. For infrastructure portfolio managers, decision support tools are becoming increasingly more useful. This is particularly relevant to railway structures as a result of their diversity and age. A thorough literature review (Chapter 2) is carried out to understand what decision support tools, known as Bridge Management Systems (BMSs), are currently available for railway bridge portfolio managers. The modelling approaches which have been used as the foundation of the BMSs are analysed (Chapter 3). Of these, the most appropriate modelling technique is selected for development of a new approach for a decision support tool. The tool comprises of a number of different modules, each with its own characteristics, data sources and features (Chapter 4). The model is presented, as well as detailed descriptions of each of the modules and how they work. During the literature review stage, a number of studies mentioned that there are external factors that affect deterioration. However, very few studies were able to pinpoint what these factors were, how much they affected deterioration and what the operational, financial and management impacts were. To that effect, a number of different factors were analysed (Chapter 5) to ascertain if they have an effect on bridge deterioration. The key factors were identified and their deterioration profiles incorporated into a probabilistic Petri-Net (PN) model, calibrated with historical data. From these, comparative model outputs pinpointing which factors affect bridge deterioration the most can be computed. Finally, simulations were carried out on the PN model to evaluate which of the factors would have the most financial effect for a transport agency. This allows bridge managers to categorize bridges in different deterioration groups allowing the definition of different optimal inspection and maintenance strategies for each group. This research has also identified that complex models often have a heavy computational burden. A study was carried out to accelerate simulations of PN models with General-Purpose Graphics Processing Units (GPGPUs)(Chapter 7). GPGPUs are composed of many smaller, parallel compute units which has made them ideally suited to highly parallelized computing tasks. The efficiency of different approaches to parallelization of the problem is evaluated. The developed framework is then used on the railway bridge PN model. The results obtained show that this method allows the combination of complex PN modelling with rapid computation in a desktop computer. A final piece of research was undertaken to perform optimisation with the railway bridge PN model (Chapter 8). This study utilised the foundation railway bridge PN model, the Local Environmental Factors (LEFs), the variability factors and the GPGPU acceleration. A Hybrid Multi-Objective Genetic Algorithm (MOGA) approach is accelerated with GPGPUs to find the optimal inspection regime to minimise both the WLCC of railway bridges and the risk of being in a poor condition. The proposed Hybrid Genetic Algorithm (GA) approach is able to accelerate the process by over 30 times compared to the traditional GA approach. The results obtained demonstrate a potential 9% reduction in overall WLCC for UK railway bridges at the same condition as the current industry policy performance. A novel Performance-Based Inspection Planning (PBIP) protocol is introduced to demonstrate where inspections should be focused to monitor bridges in areas susceptible to more severe deterioration whilst easing inspection efforts on those in milder areas of deterioration, improving operational efficiency.

It is important to monitor the wheel-rail friction coefficient in railway vehicles to improve their traction and braking performance as well as to reduce the number of incidents caused by low friction. Model based fault detection and identification (FDI) methods, especially state observers have been commonly used in previous research to monitor the wheel-rail friction. However, the previous methods cannot provide an accurate value of the friction coefficient and few of them have been validated using experiments. A Kalman filter based estimator is proposed in this research project. The developed estimator uses signals from the traction motor and provides a new and more efficient approach to monitoring the condition of the wheel-rail contact condition. A 1/5 scaled test rig has been built to evaluate the developed method. This rig comprises 2 axle-hung induction motors driving both the wheelsets of the bogie through 2 pairs of spur gears. 2 DC generators are used to provide traction load to the rollers through timing pulleys. The motors are independently controlled by 2 inverters. Motor parameters such as voltage, current and speed are measured by the inverters. The speed of the wheel and roller and the output of the DC generator are measured by incremental encoders and Hall-effect current clamps. A LabVIEW code has been designed to process all the collected data and send control commands to the inverters. The communication between the PC and the inverters are realized using the Profibus (Process Field Bus) and the OPC (Object Linking and Embedding (OLE) for Process Control) protocol. 3 different estimators were first developed using computer simulations. Kalman filter and its two nonlinear developments: extended Kalman filter (EKF) and unscented Kalman filter (UKF) have been used in these 3 methods. The results show that the UKF based estimator can provide the best performance in this case. The requirement for measuring the roller speed and the traction load are also studied using the UKF. The results show that it is essential to measure the roller speed but the absence of the traction load measurement does not have significant impact on the estimation accuracy. A re-adhesion control algorithm, which reduces excessive creepage between the wheel and rail, is developed based on the UKF estimator. Accurate monitoring of the friction coefficient helps the traction motor work at its optimum point. As the largest creep force is generated, the braking and accelerating time and distance can be reduced to their minimum values. This controller can also avoid excessive creepage and hence potentially reduce the wear of the wheel and rail. The UKF based estimator development has been evaluated by experiments conducted on the roller rig. Three different friction conditions were tested: base condition without contamination, water contamination and oil contamination. The traction load was varied to cover a large range of creepage. The importance of measuring the roller speed and the traction load was also studied. The UKF based estimator was shown to provide reliable estimation in most of the tested conditions. The experiments also confirm that it is not necessary to measure the traction load and give good agreement with the simulation results. With both the simulation and experiment work, the UKF based estimator has shown its capability of monitoring the wheel-rail friction coefficient.

Regulation requires railway energy absorbers to dissipate the collision energy and to prevent overriding. There is no industrial consensus about which energy absorbing mechanism is the most suitable for the crash conditions present in a collision between railway vehicles. There is scope for improving the existing designs or creating new concepts. The combination of two energy absorption mechanisms, expansion and splitting of cylindrical tubes, was identified as an improved energy absorption mechanism. Quasi-static and dynamic testing of scaled splitting, expansion and expansion splitting (hybrid) tubes was carried out to assess their force, stroke, energy absorption and oblique loading efficiency. In addition, the standard requires a calibrated numerical model of the energy absorber to predict its behaviour. The fracture strain of the tube and the coefficient of friction between the tube and the die are needed to build accurate numerical models. The fracture strain was measured using a Digital Image Correlation technique and a new methodology was developed to overcome its limitations. The inclusion of the fracture strain correctly predicted the deformation of the splitting specimens. The friction coefficient was adjusted until the energy absorption matched that observed during testing. Quasi-static testing showed that the force efficiency was 80%, 100% and 90%, for the splitting, expansion and hybrid tubes respectively. The stroke efficiency was measured as 77%, 44% and 70%, respectively. The energy absorption efficiency of the hybrid tubes was assessed as 11% and 40% higher than that of the splitting and expansion tubes respectively. The testing also showed that the hybrid tubes were more insensitive than the expansion and splitting tubes to the application of oblique loading. More testing may be necessary to confirm this assertion. The results suggest that the hybrid energy absorbing mechanism could become a commercial energy absorber with improved performance over the existing solutions. The validation of the hybrid numerical models showed an accurate prediction of the test results. A full-scale hybrid demonstrator has been tested and a patent of the hybrid concept applied for.

The main aim of this thesis is to investigate available techniques and develop a methodology for the fault detection and diagnostics for two engineering systems, namely railway point systems (RPS) and three-phase separators (TPS). The fault detection of the RPS was performed on the measured current from the motor of point operating equipment (POE). The method of One Class Support Vector Machines has been chosen as the fault detection model. Elastic similarity measures, such as edit distance with real penalties and dynamic time warping, were chosen to compare the data of POE operations. A combination of Euclidean distance and elastic similarity measures has been proposed in order to take into account the absolute values and shape properties of the two compared time series. The proposed methodology has been tested on the in-field RPS data. The results indicated that the fault detection model was able to detect abnormal values and/or shape of the time series of measured current. However, not in all cases these changes could be related to a recorded failure of RPS in the database. The fault detection of TPS was performed given the sensor readings of flow and level transmitters of TPS. The method of Bayesian Belief Networks (BBN) has been proposed to overcome the late detection of faults with the threshold based alarm technique. An approach to observe sensor readings of TPS in several adjacent time intervals and to update the prior probabilities in the BBN after inserting the sensor readings as evidence was proposed. The proposed methodology has been tested on the data obtained from a TPS simulation model. The results indicated that the fault detection and diagnostics model was able to detect inconsistencies in sensor readings and link them to corresponding failure modes when single or multiple failures were present in the TPS.