High Performance Computing (HPC) system for Big Data analytics

The IRRs high performance computing cluster is the engine behind the Institute’s developing expertise in data analytics and natural language processing. Working with its strategic partner RSSB and industry sponsors the facility is available to solve problems which require multi-core processing technologies – typically those found in today’s data driven systems.

The facility is based upon x86 processors and is running a modern open source enterprise operating system. It will be instrumental in the continued development of Big Data for Risk Analysis research at the IRR, allowing data to be processed in volumes not possible with desktop machines. The specification and typical railway applications for the high performance computing cluster are detailed below:

Specification:

• 5 Compute nodes comprising two 10 core Intel Xeon processors running at 2.6GHz, 128GB RAM, 512GB solid state storage
• Iceotope PETAGEN 1c Liquid cooling system
• In total the system provides 100 cores, 640GB RAM, and 2.56TB of storage
• Gigabit internal interconnects as well as an upgraded high speed connection to the University network

Example applications:

• Apache Hadoop distributed storage and processing framework, to provide an efficient storage and lookup platform for both structured and unstructured datasets such as railway close call incident reports.
• High performance textual analysis
• Apache Spark and similar big data processing platforms to provide an engine for Machine Learning and parallel Natural Language Processing

Fixed laboratory test rigs

The 1/5th scale railway roller rig consists of one bogie and two wheelsets, each is driven independently by a three phase AC motor. The motors can be controlled precisely by inverter drives. The rollers pairs are connected to two DC generators which are used to apply variable traction load between the wheel-roller interfaces.

A hardware in the loop system is built using this rig. Real-time sensor measurements are sent to a computer programme which simulates a traction controller thus sending control commands to the motor.

This rig can be used to evaluate wheel rail adhesion levels using traction motor parameters and also other research applications such as traction/braking control and regenerative braking studies.

In-Situ Measurement of vehicle and track

The IRR has a range of specialist measuring equipment used to carry out in-situ investigations and to support the validation of its numerical models. Instrumentation can be deployed on either vehicles or on track. Our main equipment is:

• MiniProf BT wheel, rail and switches and crossing measurement device (Figure 1) from Co-Mech.
• Laser measurement of rail surface (incl. squats, welds and S&C) using µEpsillon ScanControl laser head (Figure 4). Used for wheel-rail interaction studies, defects analysis and worn shape evaluation.
• Geismar track cant and gauge stick
• 5 tons track gauge restraint tester
• NI CompactDAC and CompactRIO data acquisition systems with a wide range of acquisition cards.
• Impulse hammer from light-medium weight (Kistler 5kN and 20kN) to heavy sledge-hammer (PCB 22kN) for structural vibration testing of wheels, rails, vehicle and track components (Figure 2).
• A range of accelerometers (uniaxial and tri-axials) for vehicle ride test, rail track response (Figure 3) to moving traffic at various level (rail, sleepers, bridges).
• Displacement sensors (laser and LVDTs)
• Pressure transducers for brake system measurements