Discussion Board
Group 4 – APAN Business Case |
Context
Railways are a critical part of Canadian infrastructure and the Canadian economy, transporting more than 75 million people and $250 billion of goods across vast distances each year. The rail system consists of a fleet of 2,400 locomotives pulling almost 5 million freight cars over more than 27,000 miles of track each year (RAC, 2016).i In 2015, Canadian freight trains logged over 283 billion Revenue ton-miles while traveling over 68 million miles (Ibid).ii That same year, Canadian passenger trains carried over 400 million people on 4 million miles of commuter lines and more than 4 million people on 7 million miles of intercity lines (Ibid).iii Partly due to the sheer scale of the system, over 1,200 accidents were reported in 2015 (TSB, 2016)iv. They ranged from minor to major (main-track collisions and derailments). While strides have been made in rail safety over the past 15 years, there is room for improvement.
One of the key challenges has been obtaining accurate data from accidents in a timely manner. Since 2005, the TSB has audited accident and incident reporting three times. In each instance, investigations have uncovered roughly one unreported event for every five that was reported (McDonald et al, 2013).vvi
Current Canadian reporting structures require a physical phone call and human intervention to report and record an event. Boulton (2016) reports “the number 1 cause of poor data quality is data entry by employees (57.5%)”. vii Delays of hours, days or even weeks means data essential to accident investigation is not available before an accident site is disturbed and critical evidence is lost. Equally importantly, underreporting means the veracity of the data is in question, as occur due to the manual nature of the process. Additionally, the Rail Safety Agency currently lacks sufficient data or data processing and analysis capacity to conduct regular Safety Issues Investigations (investigations that consider data patterns which suggest a developing safety issues before major consequence). Only 2 SII's have been conducted in 15 years at the TSB.
Following the Lac Megantic train disaster in 2014, efforts were undertaken to try and determine how safe Canadian railroads are, and whether changes can be accurately measured. Jennifer Winter PhD (2014), of the School of Public Policy at the University of Calgary put it succinctly: viii
“When Canadians ask, as many have in recent months, whether the rail-transport system is “safe,” they surely want to know whether the accident record is low — compared to other countries and to other forms of transport — and whether it has been improving or getting worse over time. Yet, the statistics that might provide the answers are worryingly inaccessible, sometimes conflicting, and in certain cases not available at all.”
This is not a problem related to the availability of data. The industry has a network’s system of trackside and onboard sensors that provides 10s of millions of distinct data points each day tracking the location and condition of the equipment and infrastructure.ix Big data services already support the railway industry. While there is, what Sarah White (2015)x calls “a strong culture of evidence-based decision making” in industry operations, the systems does not currently exist to support a similar approach to safety management.
In this paper, we propose designing and implementing an automated accident reporting and data sharing system.
It would be similar to the Ideagen RSSB system used in the UK (Bearfield, 2016).xi
Purpose
Improve Canadian rail safety by enabling the automated capture of real-time rail system data and providing a platform for publishing and advanced analysis of the collected data.
Objectives
• Automatically capture rail system data, particularly safety event data, from potential or actual incidents.
• Support enhanced incident triage and incident response protocols.
• Improve network wide communications to improve the velocity and veracity of safety management data
• Provide a platform for the advanced analysis and visualization of safety data and real-time system data.
• Introduce a common process for tracking, managing and investigating safety events as well as for tracking and managing local actions and recommendations.
• Define industry-wide key safety performance indicators (KPIs).
Assumptions
• Industry cooperation in exchange for reduced reporting costs and shared financing of rail network control costs, thereby bringing rail industry infrastructure overhead in line with trucking industry.
• Government financed and operated by independent investigation agency (UK model)
• Regulator framework reconciled with data tracking and reporting requirements (automated data capture).
Acceptance Criteria
• 25% reduction in reported accidents and incidents over 5 years, compared to baseline.
• 99.9 % system reliability.
• Proof of concept demonstrates proposed system more reliable and less costly to operate than existing system.
Data Requirements
Anticipated Data Sources
Locomotive on board equipment telemetry and health data | Government legislation require all locomotives to be capable of collecting an storing structured data on information such as speed, distance, train air brake function, throttle use. Companies such as GE (i.e. Trip OptimizerTM) have augmented these systems with sensors, storage and satellite/cellular systems capable of up linking data to servers in for continuous monitoring. When serious operational events occur, such as a collision or derailment, or even a train separation, data is automatically captured and sent to support event analysis. It is not automatically sent to the TSB. |
Rail car location and health monitoring | Structured and unstructured- At least 40,000 rail cars currently employ Industrial Internet of Things (IIoT) sensors on cars that, at a minimum, capture GPS location, date/time, and accelerometer data in 3 dimensions. Some versions can capture video or acoustic signals. GE and IONX are currently in a partnership to use this data to produce the “smart train.” |
Shipping container tracking devices. | Increasingly, companies shipping their goods in containers are using IIoT technology to track the movement and handling of their goods. |
Track-side detectors | Wayside systems capture information from a passing train, such as wheel and wheel bearing temperature, signs of dragging equipment, wheel shape and surface damage. These used to be stand-alone systems, but increasingly are captured through RailInc., an industry-wide equipment health and maintenance initiative providing key data to railway companies and car owners. Event triggered data access. |
Rail Traffic Control Centre Data and signal system downloads | Mainline railway operations are mostly conducted on railway track that provided signal information back to a central control center (think, air traffic control center) where it is used to direct the movement of trains. These systems can capture instances where track is damaged, causing signal system failure, as well and instances where trains move outside the prescribed limits of where they are allowed to operate, increasing the risk of a train collision. |
Video and audio recordings | Regulations required the capture of certain radio and phone communications between traffic control staff and other operating employees. Increasingly, video cameras are used like surveillance system cameras to help protect yard movements and monitor commuter railway station platforms and this data is often recorded and used internally by companies. |
Locomotive forward facing video | (un-structured). Most freight, passenger and commuter railroads equip locomotives with forward-facing cameras that can capture crossing or trespasser incidents. |
Locomotive voice and video recordings | (un-structured) There is a strong push by industry and regulators to establish standards for the collection of voice and video recordings of the crew operating a locomotive. Pilot tests are ongoing and it is likely this new source of data, and systems to monitor the safety implications of that data, will be coming online. |
Twitter / social media | (un-structured) When major accidents occur, they are accompanied by a surge in digital communications – twitter, text, video, voice. |
RailFans – video, photo capture | There are websites dedicated to the activities of rail enthusiasts who videotape and photograph railroad operations. In the past, they have captured safety significant events. |
Phone records | The railway industry relies on heavy equipment contractors who specialize in dealing with the results of a railway accident. These contractors are usually called before the TSB. Monitoring the phone lines to these contractors, together with other data, would give indication of the seriousness of an accident. |
Government weather and environment data | Environment Canada severe weather data, Resource Canada seismic data and other surface geological and hydrological ground hazard information. Transport Canada track inspection vehicle track defect records. |
Track geometry and ultrasonic inspection data | Industry regularly inspects track for geometry and surface changes that can destabilize a train, and defects in the rail that could cause the rail to break. |
Crew work-rest schedule | Fatigue monitoring capacity. Perhaps, eventually, alertness sensor information. |
Positive train control (PTC) | Automation of train control in next 5 years will dramatically increase amount of data gathered and used prescriptively in industry. |
Anticipated Data Volume
Petabytes of structured and unstructured datat scanned. Only anomalous triggers stored. Over time, exabytes may be scanned and stored.
Frequency of Analytics
Access existing real time analytics monitoring structured (IIoT, wayside networked detectors and signaling system) and unstructured (video, audio, photo, select contractor hotline reporting phone number) data sources to capture data anomalies consistent with train derailment, collision events or risk of collision events.
Daily and weekly analytics to generate and review KPIs, capture trends and detect new patterns.
Data Timeliness and Reason (refresh periodicity, real time)
System requires 24/7 redundancy. Anomalous events need to be captured and stored for trending.
Archiving/History Requirements
Currently, there is no automated system for entering data into the TSB internal Oracle database. Events are reported by email or phone, and then entered by hand. It is generally accepted that the veracity of data in the system is not adequately validated.
Under RSSB model (used in the UK) approximately 75,000 events are captured annually. In Canada, currently, 2000 events captured through phone-in reporting. Data storage capacity would have to increase by roughly 3000%, eventually growing to incorporate cloud-based function.
Other/Optional Elements for Consideration
System requires 24/7 redundancy.
Security Requirements
System would tap into company proprietary data streams. Data has business value, and reporting has privacy requirements, so data must be gated from sources (suppliers cannot know what the other is supplying), and must be secured from inadvertent or directed efforts to hack into database. Open public data initiative should provide data of sufficient volume, velocity, and veracity to support public safety initiatives similar to Open.gov in the US.
iRailway Association of Canada. (2016). Rail Trends 2016. Retrieved from http://www.railcan.ca/publications/trends
ii Ibid
iii Ibid
iv Government of Canada. (2016, July 19). Statistical summary - railway occurrences 2015 - transportation safety board of Canada. Retrieved February 10, 2017, from Transportation Safety Board of Canada, http://www.tsb.gc.ca/eng/stats/rail/2015/sser-ssro-2015.asp
vTSB subpoena’s on CN Rail’s records for 2001-2007 resulting in 1800 additional reportable occurrences. Following the Lac-Megantic disaster, about 60 additional Montreal Maine & Atlantic occurrences were uncovered in company records, and a further audit of two Class 1 freight railways following news reports again identified hundreds of un-reported occurrences.
vi McDonald, J., Noel, B., & Files. (2013, December 9). CN rail did not report 1, 843 accidents. Retrieved February 10, 2017, from News, http://www.cbc.ca/news/canada/tsb-says-cn-rail-failed-to-report-hundreds-of-derailments-collisions-1.2451186
viiBoulton, C. (2016, February 5). Disconnect between CIOs and LOB managers weakens data quality. Retrieved October 24, 2016, from http://www.cio.com/article/3030249/business-analytics/disconnect-between-cios-and-lob-managers-weakens-data-quality.html (Links to an external site.)
viii Winter, J. (2014). Safety in numbers: Evaluating Canadian rail safety data. SSRN Electronic Journal, 6(2), 1–10. doi:10.2139/ssrn.2419723
ix Footnote
xWhite, S. K. (2015, November 10). Study reveals that most companies are failing at big data. Retrieved November 16, 2016, from http://www.cio.com/article/3003538/big-data/study-reveals-that-most-companies-are-failing-at-big-data.html (Links to an external site.)
xi Bearfield, PhD, MSc, BSc, CEng, FSaRS MIET, G. (2016, October). Developing dynamic safety management capability in the rail industry.