Not Quite, Murder on the Orient Express

It does not take murder and a bunch of Hollywood A-listers to excite us here at Informatica. Providing safe passage, a smooth and profitable ride on rails still does though.

If you work for a railway or your organization has USD276 million a year to maintain all your physical assets, would you like to know how to boost on-time performance, safety and save a third of your annual maintenance budget using cutting edge data management techniques?

If so, buy a ticket to this matinee right now.

IIoTWell, as luck would have it, an operator in Europe operates tens of thousands of miles of tracks and supporting network infrastructure for more than twenty passenger and cargo rail operators. They are charged with maintaining a profitable and safe infrastructure, larger than the total rail network in China. Close to 200 years after George Stephenson built and operated the first public steam railway from Stockton to Darlington, this organization is charged to maintain safe and commercially performant rail operations.

Defunct tracks or other assets reduce uptime and therefore can cause delays, which in turn can inflict severe penalties from its regulator in the neighborhood of $42-$84 million.

The typical problem in this scenario – and we see the same in the power utilities space – is the historic growth of asset management silos. No matter what someone tells you; if they work for a company, which has been in business for 50 or more years in some form, there never is just one asset master (even despite a recent ERP implementation). These silos of information grow over time and typically receive little to no affection as they are considered vital, yet just cost-of-doing-business. They are never considered investment-worthy, critical differentiators or even profit enhancers.

So, what are we to do here? Essentially, this organization has three major options:

  1. Do nothing and meddle along
  2. Implement a central asset registry (likely IBM Maximo)
  3. Virtualize its 8+ asset registries using a central synchronization hub with minimal source system intrusion

Obviously, doing nothing would not fix their recurring multi-million-dollar fines. Their internal analysis shows that implementing a central asset registry and retiring the 8+ legacy ones is too costly as well as too risky as they cannot be down for a moment due to the ongoing maintenance operations. Option 3 it is then.

The Plot Thickens

The environment basically strings the 8+ asset registries together and matches nine different asset classes (signals, rail, stations, rolling maintenance stock, etc.) together based on location (GIS), specification and asset type characteristics. As it turns out, what is an elongated piece of metal strapped to wooden logs on a gravel bed to you and me is quite a bit more to the rail track connoisseur. There are actually 28 different type of tracks and they make up between 60-80% of the annual maintenance budget.

For each type we would use 5-6 different match rules and based on best-fit given the characteristics and trust level in the respective sources, one merge rule is selected to conflate the various track profile existences into one.

Just to give you an idea about magnitude, there are just over 1.5 million track asset records with somewhere between 400-600 attributes. Mind you that the average asset data master runs at maybe 20-30% of that.

Once the clean track asset profile is in the hub, and this could be a relational or Hadoop environment, it can then move onwards to an ODS (Hadoop based as an option) and enriched with sensor (IIoT) data around track tilt, bounce and track parallelism. These three factors were determined to be good indicators of future maintenance need. In the olden days before this project, maintenance was neither condition-based or predictive but purely interval based. Once again, this is something we are very familiar with as I outlined in a prior post on the utility sector.

The track condition data is collected via optical sensors mounted on the bottom of trains scanning petabytes of data in 50 cm intervals. Once certain tolerances are exceeded, they are written into the ODS at the appropriate geolocation and performance characteristics of a 50 cm (half a yard approx.) piece of track. This in turn could be part of a 2-yds to 2-mile long welded piece of track so its larger context is quite important from a planning and budgeting perspective.

Armed with the information on the exact location, type, issue and suggested remedy, the maintenance planning team can then prioritize their work based on severity and traffic considerations. More importantly, the team can also use a special linear visualization tool to identify and understand any lower priority, yet close proximity problem track areas to do additional work while on site earlier so as to keep efficiency high.

The point of this post is to peel the onion a bit more around IIoT use cases for the more inquiring mind. I hope it is apparent now – no matter what assets you manage – that investing in your physical assets and infrastructure is of paramount importance if you operate an asset-intensive business model. Adding new data sources (sensors) quickly, combining them with clean core data (asset master profiles) and in the future allowing artificial intelligence to further optimize failure prediction, will provide not only superior safety to your customers but also generate financial rewards.

In summary, most of the required raw ingredients in terms of technologies have existed for a while but the secret sauce lies in the ability to integrate them and most importantly, change leadership’s mind about how to compete in an increasingly more competitive and democratized information age.