Green Engineering: Rail maintenance system advances with digital embedded design

by Anders Norlin Frederiksen (Analog Devices) and Marco Schmid (Schmid Engineering) , TechOnline India - September 30, 2008

This article describes a Blackfin/LabVIEW-based rail-maintanance system that measures, locates and fixes rail defects while achieving the real-time behavior, cost-effectiveness and robustness required in that harsh environment.

Public transport by rail or tram has over the last decade become a popular way of transportation. The number of passengers taking advantage of a comfortable and safe ride is constantly rising which calls for higher train speeds and shorter stop intervals, exposing the rail- and tramways to extreme mechanical stress.

Unavoidably wearing out in time, not only annoying, but also dangerous, defects start tweaking in. A new and systematic maintenance solution (Fig. 1) [1] supports the monitoring and maintaining for rail- and tramways.

Fig. 1: A systematic rail maintenance concept includes measuring, locating, planning and fixing rail defects.

Analog Devices Blackfin processors and National Instruments LabVIEW plays a central in the system, securing all measurement and field-data correctly to be stored for immediate action. The result is clear, much longer operation times between exchanging rails, ensuring the public transport service a more economic and successful evolution than ever before.

Railway Tracks "Under the Hood"
When new rail- and tramways are laid out, high quality assurance verifies correct track positions prior to concreting. Some time after installation, several defects start to tweak in to the rail parameters (Fig. 2) during the daily operation.

Fig. 2: Rail parameters are divided into track geometry, longitudinal profiles and cross-sections.

This is due to the mechanical contact between the wheels and the rails in relation with a highly complex dynamic spring-mass model ranging from the train chassis to the railway underground. The defects, their critical parameters and tolerance windows are classified by railway engineering standards [2]...[14]. It's the goal of this rail maintenance program to deal with the irregularities but keep them to acceptable levels.

Rail Track Geometry
The track gauge is the distance between two rails and responsible for the so-called "sinusoidal ride" of the train. This keeps the spot where the wheel and rail meet constantly moving to minimize wear-out.

Variances in the track inclination can make passing trains shake and shudder. Mostly caused by giving way of the railway underground, inclination defects can also be kicked off by surface irregularities such as corrugations and holes. Systematic inclination profiles are however necessary to minimize accelerating forces to the passengers when a train is riding in and out of a curve. A correct track-to-track distance prevents any chance of collision when trains are crossing at high speed.

Longitudinal Surface Profiles
Cracks and breakouts are among the most feared since they can lead to catastrophes such as derailing. Corrugations on the other hand are wavy irregularities with a characteristic wavelength between 20 to 100 mm and are annoyingly noisy when their amplitudes exceed 0.05 mm. From 0.3 mm peaks however the vibration can leave irreversible damage to the railway bed.

It's also in their nature to move along the rails and scientists still debate where they originate. Single holes are mostly generated by turning or jumping wheels and follow the mathematical equation of a polynomial. They're responsible for the sudden bumps on a tramway ride. Regular bumps that are often experienced on older railways are due to the welding interfaces of the 18-m railway sections.

Cross Sections
The head geometry of a newly installed rail follows an exactly calculated contact geometry which optimizes the critical wheel-to-rail interface. The shape consists of tangential lines and specific radii allowing the wheel to economically and smoothly roll off with a safe horizontal guide.

{pagebreak}Measure the Rails
The key requirement for systematic and target-oriented rail maintenance is a covering knowledge about the current state of the rail- or tramway network's geometry (see Fig.2). This is achieved by a smart measuring strategy that combines odometer results (distance measuring), track geometry, longitudinal profiles and cross sections with exact GPS locations.

All these parameters are acquired by mobile metering devices or complete measuring vehicles. Initiated and pre-processed by Analog Devices Blackfin processors, the measurement data is finally transferred into a high-level analysis software that allows post-analyzing and pinpointing the measurements and defects on a digital map (Fig. 3).

Fig. 3: Measurements are combined with GPS data to pinpoint them in Geographic Information Systems (GIS).

Track Geometry
The rail gauge is measured using no contact inductive sensor principles with accuracies in the 0.01 mm range. Software based FIR low-pass filters suppress high-frequency noise while subsequent moving average filters ensure that no "pseudo-peaks" are occurring in a result that is expected to be continuous.

A similar approach is applied to the inclination sensor that operates like an electronic water-level with an angular range of ±10° and an accuracy of < 0.025°. Due to the physical principle, results are only valid in a certain frequency range, typically below 1 Hz.

Measuring the track-to-track distance requires a set of complex and computational demanding floating-point algorithms to finally deliver the relatively simple result of the absolute horizontal and vertical distance (Fig. 4).

Fig. 4: Measuring the track to track distance (X/Y) demands for high-performance digital signal processing algorithms at runtime.

A high-precision laser beam which is attached to the side of a vehicle, wobbles ±5° within a distance range of 1 to 5 m under the control of a Blackfin processor. The profile of the neighboring rail which is expected within this scanning sector, is low-pass and median filtered and transformed from the polar into the cartesian coordinate-system. After applying some processing like vector-rotation and resampling, the profile passes through a pattern matching algorithm.

The goal is to find the exact vector to a characteristic geometric feature within the realhead. Due to the many obstacles like rocks or grass that are found on railways, this vector finally passes through a plausibility checker and a tracking algorithm to provide reliable and valid results. All this is done in a 5-Hz loop under real-time conditions.

Longitudinal Profiles
High-speed eddy current sensors record both rail surfaces with micrometer accuracy (Fig. 5).

Fig. 5: Longitudinal rail profiles are acquired by no-contact eddy current sensors, pulsed by magnetic encoders.

A linear encoder processes signals from a magnetic ring that serves as an odometer and as a trigger for the AD sensor converters. This signal then goes through a FIR (Finite-Impulse-Response) filter with a bandpass topology reducing the spectrum to the characteristic wavelengths. On top of the surface profile, also metallurgical irregularities such as partial hardenings and welding points are recorded.

Cross Profiles
Laser technology is todays state of the art non-contact measuring principle to get the exact cross section of a railhead. Depending on the required accuracy or capturing speed, either traversing laser beams or laser "curtains" (Fig. 6.) are used to do the job. Raw profiles are linearized, scaled and spike filtered in real-time.

Fig. 6: Rail cross profiles are captured by high-speed laser scanners.

Older Technology - Metering Devices
Until some years ago the maintenance staff used many different metering devices identifying cracks and variances on the rails. Each methodology specialized in recording one specific rail defect and almost all these mechanical methods lacked precise and reproducible results.

In recent years industrial solutions providers like Schmid Engineering have taken advantage of embedding advanced processor technology and state of the art methodology into design. Advancing these methods into the railway infrastructure business gradually empowered mobile and multifunctional rail measuring by smart metering devices.

Rail monitor devices (Fig. 7) uses state of the art measurement technology to simultaneously define the cross profile of the rail, the head height, track gauge, inclination, depth and ambient temperature, detected and logged at any specific location.

Fig. 7: Rugged environments and tight schedules demands for light, easy to use and productive metering devices.

All key characteristics are processed and visualized on-site and stored to removable memory. The Railsurf sled (Fig. 8) continuously monitors and records longitudinal track parameters as an operator or a vehicle pulls it along the rails.

Fig. 8: The RailSurf sled, driven by Blackfin processors and LabVIEW Embedded, records longitudinal wavy irregulatities. A GPS receiver and inclination sensor is built into the operator panel.

It carries several sensors, mapping problems as corrugations, holes, cracks, variations in rail gauge and inclination. The resulting information can go onto removable memory or wirelessly transmitted to any operator interface.

Blackfin Processor as the "Heart" in the system
The Blackfin Processor as the "heart" of all these test tools empowers the convergence of MCU and DSP technology through offering dynamic power management for any given battery operation. The MCU part conveniently interfaces with scalable I/O like laser scanners, analog and digital sensors, keyboards, TFTs, batteries/fuelgauge and removable media. The DSP part is dedicated to advanced digital algorithms like filtering, FFTs or determination of the geometric residuals or other demanding computational tasks.

Recent advancements in graphical system design by LabVIEW Embedded offers a direct programming model of any Blackfin processors with its high level block diagram and dataflow-oriented language. This high-level approach with ready-to-use mathematical analysis blocks and graphical multitasking moves functionality to the next level of digital embedded design.

Measuring Machines
A multifunctional vehicle that is driven by a set of five interlinked Blackfin processors is able to record rail parameters up to 10 km of a railway section with 5-mm point-to-point resolution.

Blackfin #1 allows user interaction over a keyboard and two TFT monitors. Blackfin #2 records track geometry and longitudinal profiles at high speed and embeds GPS information into the measurements which is received by Blackfin #3. Together with cross-sections that are captured by Blackfin #4, all the data is finally streamed to Blackfin #5, which stores the huge amount of data in RAM buffers to be eventually saved to binary files on removable media.

{pagebreak}Locate the Defects
The acquired measurements are now fed into a common software platform that links track geometry, longitudinal profiles and cross-sections with GPS locations and odometer information. Realized with LabVIEW and its toolkits, this platform serves as a common data exchange and analysis pool. It interfaces to a variety of measurement devices, vehicles and maintenances machines.

Smart filters applied to the measurements function similar to an X-Ray by locating the most critical rail defects. The result is a true digital representation of the whole rail geometry (see Fig.2). This essential information is now directly usable for actions like rail repair or exchange. The final datalog is again wirelessly connected to external databases and CAD software to transfer the results into any customer's IT environment.

LabVIEW filters find defects
Smart LabVIEW filters sift through longitudinal data to find symptoms of interest. Corrugations get detected through a Fast Fourier Transform Analysis (FFT) watching for the characteristic wavelengths in the longitudinal profile.

Holes are tracked by comparing the measured profile with memorized patterns and the simulation of the mechanical rail-wheel contact. Cracks show significant transients so they can be detected by differentiating a moving data window. Finally, unique vibration patterns in the inclination profile are located by continuously running and evaluating analytical models.

The resulting symptoms are also fed into correlating "super-algorithms." Here either the information is reduced even more or additional high-level information is extracted out of the measured data. An inclination symptom, for example, is meaningless and therefore rejected if there's no related signal peak on the rail surface. However, a cross profile indicating a significant wear-out with a crack in the longitudinal error will trigger an alarm.

The main concept behind the evaluation of rail cross-sections is comparing a measured profile with a reference. Algorithms based on vector mathematics and stochastic methods align and overlay the two profiles to allow derivation of critical characteristics. Vertical and perpendicular residuals directly indicate wear-outs (Fig. 9).

Fig. 9: Smart cross profile analysis algorithms take advantage of the high-performance Blackfin Processor performance to indicate irregularities in real-time and on-site.

Other parameters include the remaining head height, a correct and well shaped rail radius (Fig. 10) or the gap of an active, closed switch.

Fig. 10: Determining the rail radius called for complex mathematical functions. Thanks to algorithm engineering with LabVIEW, the concept has been proven in only one working day.

Keeping the tolerances of the latter is a key requirement for high-speed trains passing the switches without the danger of de-railing. That's why rail operating companies now focus on thoroughly monitoring the switches. Rail engineers can tweak filter parameter tolerance windows to separate "pseudo-warnings" from true rail defects that significantly influence either the passengers comfort or safety.

Pinpointing "defects" on a Digital Map
GPS data that is embedded in the located defects allows pinpointing them on a digital map (see Fig. 3). This geographical information adds significant knowledge and new context with regards to railway "hot-spots" such as tight curves, switches or stations. This "Easy-GIS" (GIS = Geographic Information System) has been realized with the image processing features of LabVIEW.

An existing bitmap of the region of interest, e.g. a city, is broken down into single tiles, each given the exact map coordinates. As the rail engineer is browsing through the set of defects, LabVIEW continuously loads the according tiles from the hard-drive into the memory and assembles them to a single JPG image. This image is then copied into a LabVIEW plot chart indicator and overlayed with a digital cursor at exact locations of the defects.

Distribute Results To Other Applications
The results are finally transferred to and from superior applications. Geometric profiles of critical defects such as wear-outs and holes can be exchanged with standard CAD Systems for further analysis. This is achieved by the DXF (Drawing Exchange Format) file format.

Connection to external database management systems is established through ActiveX Data Objects (ADO) which uses Universal Data Links (UDL) for the connection type and path. A set of high-level VI's allow the data platform to perform the most common database tasks such as addressing tables and exchanging data.

The VAG transport corporation Nuremberg maintains a matrix of predefined and critical locations in a MS Access database which is continuously screened for variations. As soon as some hot-spots exceed a tolerance window, an electronic maintenance plan is created and deployed to the measuring devices in the maintenance machines.

The maintenance concept at Zurich Public Transport (VBZ) relies on a commercial GIS tool with a built-in MS Access Database. All infrastructure elements including rail sections, stations, switches, etc are listed and can be visualized on a geographical map that represents the whole tram network of the city at the push of a button. Similar to Nuremberg, the state of the rails is continuously monitored as a vital part of a short- and long-term maintenance concept. The LabVIEW platform connects to this GIS tool by the means of ActiveX and mechanisms.

{pagebreak}Solving the Problem
The resulting maintenance plan fed back from the IT environment is downloaded into the maintenance machines as quality set points. A dual Blackfin processor architecture supports the team by fixing worn out or defected rail sections fast and systematically. This is achieved by several iterative grinding runs to bring the rail back into its original shape.

One of the Blackfin processors provides a multifunctional keyboard, a visual of the rails on two TFT monitors and removable memory to the personnel. Two laser scanners continuously capture snapshots of cross profiles at 20 Hz and transfer the data online to the CPU over a CAN network. The processor then calculates the deviation to a reference profile and forwards new set points to the underlying grind unit which is controlled by the other Blackfin processor.

This grind unit consists of a total of six independent grinding pots. Each offers three degrees of freedom with actuators based on a hydrostatic principle. At first the pot moves horizontally either to the inside, outside or middle of the rail head. Then it rotates to the worst case deviation and finally moves down until it touches the railhead to start removing the material.

The Blackfin processor controls each of these 18 movements simultaneously by applying PWM signals to the valves that control the hydrostatic actuators. Additionally, six rotation sensors, six translation gauges, 18 no-contact position switches and six pressure sensors are continuously monitored during this positioning process. While this took minutes using a traditional approach, the grinding pots are now placed automatically within seconds.

Finally the grinding pot starts to remove the excessive material (Fig.11).

Fig. 11: Electronic maintenance plans are deployed to maintenance machnes to solve the problem on-rail by grinding.

Secure and sturdy housings protect the electronics and the sensors from flying sparks, aggressive dust, humidity and heat. After the grinding process, the quality is assured by loading up a set of profile measurements back into the IT environment using removable media.

About the authors:
Anders Norlin Frederiksen is World Wide Industrial Marketing Manager at Analog Devices. Marco Schmid is Senior Engineer at Schmid Engineering, Switzerland.

1.  Vernetzte Schienenmess- und Schleiftechnik, "EI-Der Eisenbahningenieur" Ausgabe 6/2007
2.  Bahnanwendungen - Oberbau - Abnahme von Arbeiten, EN13231-3 --NORM
3.  Katalog der wichtigsten Schienenfehler in Gleisen und Weichen, Richtlinie 821.2017.Z01DB
4.  Prufung der Gleisgeometrie mit Gleismessfahrzeugen, Richtlinie 821.2001
5.  Prufung des Schienenkopflangsprofils, Richtlinie 821.2008
6.  Stossluckenprufung, Richtlinie 821.2009
7.  Zulassige Abnutzung der Schienen im Gleis, Richtlinie 821.2011
8.  Langwellige Gleislagefehler messen und erkennen, Richtlinie 824.0520
9.  Messeinrichtungen und Handmessgerte, Richtlinie 824.0540
10. Bearbeitung von Weichen, Richtlinie 824.4016
11. Schienenbearbeitung in Gleisen, Richtlinie 824.4015
12. Neuschienen bearbeiten, Richtlinie 824.4010
13. Schienenbearbeitung planen, Richtlinie 824.4005
14. Schienenbearbeiten Grundlagen, Richtlinie 824.4001

Related links:
Under the Hood--Extreme: Monitoring Gas Extraction in the Norwegian Sea
Six-legged robot spider does the limbo
Graphical tools for rapid design, prototyping, and deployment
LabView module supports ADI Blackfin Processors


blog comments powered by Disqus