Seeing the lights
Creating a 'virtual railway' allows signal locations to be optimised with minimal disruption to traffic, simultaneously reducing costs and producing better results
Ted Stephens, Chris Angus and John Bryant, Bentley Professional Services
AN EFFECTIVE signalling system must bring together a number of different elements: electrical and electronic components, software and logic, the signal layout, and the interactions with drivers.
Years of study and analysis of hardware and logic have delivered very reliable signalling equipment. More recently, events such as the 1999 collision at Ladbroke Grove in west London have highlighted the need to focus on signal location, sighting distances and the way in which signals are interpreted by drivers. This is challenging, as traditional techniques for choosing the positions of signals have become harder to use.
Safety regulations make it difficult for staff to gain access to the track, and increasing traffic means operators wish to minimise disruption. Route learning is difficult because a new layout cannot be driven until it is completed, and even then it may not be possible for a driver to pass along every possible route.
Virtual railway model
The rail industry has thus had to devise new ways of establishing signal sighting, and communicate final designs to train drivers. An approach which has been used on the Channel Tunnel Rail Link, Manchester South project and the current Portsmouth resignalling uses a three-dimensional computer model of the railway, which can be used to place signals by parametric design rules and then perform objective assessment of sighting, aiding subjective assessment within the real environment.
The models are generally based on a number of datasets. A static element consists of the terrain, buildings and earthworks, which rarely change. There is also a semi-static model, comprising the rails, overhead line and other features that can be updated on an intermittent basis, requiring some data changes. Finally, there is a dynamic model, which holds the signals themselves and is automatically updated by the signal-sighting software as each signal is placed and design decisions evolve.
These elements combine to form a complete model of the railway, which can be analysed and viewed at will. The output can be generated in a variety of forms depending on the user's needs, including drawings for design engineers wire line diagrams, static images and animations for external users and interactive viewers for subjective assessment.
The most important feature is that this should be a seamless process, using a single data model generally known as a virtual railway. This requires a variety of software components:
- a three-dimensional string modeller for terrain, track geometry, earthworks and overhead line
- a solid modeller for structures and buildings
- a rendering engine to create high-quality still images and animations
- a signal placement and sighting tool
- a viewer for quick visual assessment of the design
- a simulation system for communicating the design to the drivers.
All of these components should work from a single data source to guarantee precision and currency of information. The models are built in a manner that simplifies the change process, but there is still the problem of obtaining all the data needed to build the models. A wide variety of sources can be tapped, depending on the precision required.
It is reasonable to use data from a national mapping agency as the basis for the terrain model. While this has known limits of precision, it has the advantage of being available for the whole rail network. It can be used for an objective analysis of signal sighting if suitable tolerances are allowed. Where greater precision is required, more detailed data can be obtained using aerial or ground surveys.
Overhead electrification models can be constructed from as-built records or from on-train measurement systems, and the shapes of tunnels and bridges can be established from gauging records or scanning surveys. Information on the existing signalling layouts is generally derived from records or from survey data.
A number of different software tools are used to meet the signal-sighting requirements, including placing signals, performing calculations, creating 'flight lines' for the driver's eye position as the train moves along the routes, and interaction with the viewer.
The placement tool allows the designer to create signal objects in the 3D model and configure them. The tool allocates an identity, composes the signal from components, places the signal in the model, and sets signal direction and dip. Once these have all been assigned, the signal can be placed into the model for immediate viewing.
The user can change the signal or any of its attributes at will, immediately seeing the effect in the viewer. It is simple to adjust the placing of the signal to avoid any problems such as read-through or obstruction by other equipment.
The performance calculations tool allows the user to perform intervisibility calculations based on the location and direction of signals, obstructions, and the driver's eye flight line. The tool creates rays for line of sight, calculates obstructions and the 7 sec and 4 sec positions required for first glimpse and continuous sighting. It can also identify possible read-through locations, establish a reserved signal volume, and create signal-sighting forms. The line-of-sight calculations take into account the conicity of the signal lamp lens and the viewing angle for the driver's eye, constrained by the viewing angle of the cab window.
An obstruction calculation is performed for features rising from the terrain (typically buildings and masts) or dropping from above the line of sight (bridges, tunnels and electrification). The view of each lamp within each signal is checked for each point along the flight line for the driver's eye, typically every 10 m.
Lines showing the rays from the signal to the driver's eye position are generated in 3D and can be viewed in plan or wire-line perspective. Based on the 3D rays, a reserved volume can be established for the signal, into which no obstruction can be placed without reference to the signalling engineer. The final product of this process is the signal-sighting form, with the basic details for the signal filled in automatically.
In normal operation, the static model of the surrounding buildings and the semi-static model for track and overhead equipment are read into the viewer, and the signal design model is then passed dynamically. As signals are created, modified, or deleted in the 3D model, the interactive view is automatically updated.
Passing the data to a full rendering engine can create still images and animation. A ray-tracing functionality allows the user to generate images with shadow and reflectance, producing realistic images which are suitable for presentations to regulatory authorities and the public, as opposed to the engineering-assessment quality output from the interactive viewer.
The downside of this is the computing power required. Each frame of an animation can take a minute to generate, and a smooth animation requires a minimum of 25 frames per second a 30 sec animation sequence can take many hours to generate, and any changes will usually require a re-run of the animation. So although this method does not lend itself to the interactive assessment of signals, it is ideal for display of the completed work.
High-quality still images can be generated using the rendering engine. It is possible to include shadow, reflectance, fog, cloud, bright sunlight, and many other characteristics. Each of these added features requires additional computing time, but a very complex still image will only require a few minutes to render, rather than hours.
One of the key features of this approach is the speed with which it is possible to achieve results. The interactive viewer allows the immediate display of anything that has been built to date, and it is possible to start building models around a local area of interest and then to expand this at will, unlike animation where the modelling of the whole project must be completed before any animation can occur.
The data is generated in an industry-standard format widely used for visualisation and simulation systems, so it is possible to upload the models into a simulator for route familiarisation. Alternatively, the interactive viewer can be used in stand-alone mode, which allows the user to select various train paths and set the signals as required. Either approach builds on the same base dataset used for the design of the signals, without the need to reconstruct the data in a new system.
The benefits of this approach are many. From a practical and safety perspective, it will reduce the number of site visits, and thus possessions, required and the need for staff to go onto the track. From a quality assurance point of view, this approach gives a continuous and traceable source for the data on which any decisions are based, helping to eliminate the possibility of out-of-date information being used.
From the cost side, the benefits are substantial not only in time saved by reduced site visits (though these will still be necessary to confirm that the computer model is correct), but also from reduced reworking as most of the location problems will have been resolved during the design phase.
This approach to signal sighting has proven to be a great success. For the first time it is possible to link the design process and objective analysis with the subjective analysis and route learning available through the interactive viewer.
- CAPTION: A wide array of objects is brought together to produce a virtual model of Stockport in the UK
- CAPTION: A virtual model of Heaton Chapel was used to plan signalling work near Manchester
- CAPTION: Software tools show designers a driver's sight lines to the signal