Predictive Maintenance

Mar 17, 2016

Transportation companies use real-time analyses of machine data and external data sources for predictive maintenance and predictive operations to free up capacities.
Die Logistikbranche wird durch smarte Lösungen zunehmend revolutioniert. Einige davon präsentiert T-Systems auf der CeBIT 2016.
Commuters on local trains in Cologne need a lot of patience. For around six weeks in March and April 2016, the railway company is renovating and renewing the tracks and switches in the entire Cologne network. Delays and cancellations cannot be helped. Cologne is one example why companies transporting people and goods are working hard to optimize the logistics workflows of a network infrastructure that's pushed to the limits of its capacity both in rail transport and on the roads. How can railways transport more people and goods as punctually as possible with the same resources? And how can freight forwarders transport more goods on the road and increase the utilization of their fleets?

Predictive maintenance, operations and analytics

Studies show that 35 percent of initial train delays are caused by malfunctions and defects in their rolling stock – locomotives, train sets, freight cars and passenger cars – and infrastructure components. Reducing this rate would improve workflows and lower costs. The magic word is predictive, in other words: developing forecasts on the basis of real data. Union Pacific Railroad in the United States already relies on predictive maintenance of its locomotives, saving around 100 million dollars per year, as the company claims. The locomotives continuously send status data for their motors and critical components. Special software analyzes this data and reports malfunctions before a breakdown actually occurs.

Combining machine data and external data sources

Predictive operations go one step further, because the data analyses also include information that is not related to the machine status. For example, they combine environmental data, GPS coordinates and geographical data with the train data and detect faulty measurements. If the system identifies hazards, it defines adapted driving behavior, thus preventing emergency braking and reducing the number of delays.
Predictive analytics, which is based on artificial intelligence methods, delivers the most valuable information. Mathematical algorithms forecast the future behavior of systems on the basis of historical data – comparable to an ECG which compares normal readings with measured data and detects risks. On machines, this method can identify data that typically occurs just before something goes wrong.     

Better information for travelers

Predictive analytics methods can also be used to provide travelers with better information, because they want to know whether an unforeseen event will change their departure time and result in an unexpected waiting time. At CeBIT 2016, T-Systems is presenting a data analytics solution that offers scheduled transport service providers and their customers automatic information in real time about the departure times of long-distance trains and buses. T-Systems' data analytics platform calculates in only ten seconds the anticipated effects of changed train arrival times on connections for the entire network of a European railway service provider operating up to 40,000 trains per day.

Analyses with SAP HANA and Hadoop

The service continually compares theoretical timetable information with the current transport situation and the regular status reports from the individual forms of transport. This data is then used as the basis to forecast the probable arrival time and, simultaneously, the effect of this on potential connections. Passengers can in turn receive exact information in real time up to 90 minutes before their planned departure on the actual departure time. T-Systems uses cloud-based SAP HANA and Hadoop for its data analytics. The services are provided from certified high-security data centers that comply with Germany's particularly strict data privacy protection laws.