There’s an urgent call to Service: A train has come to a halt on an open line and is blocking an important rail connection. The vehicle driver suspects a malfunction in one of the drive motors, cites an error number reported by the on-board system, and awaits a concrete recommendation: whether it’s safe to resume the journey, what train personnel need to do, or whether they’ll have to be towed. “This type of situation is a case for real experts,” says Thomas Müller from Siemens Mobility. “The time pressure is extreme and the decision about what to do has to be based on numerous facts and boundary conditions. A fault is like a logic puzzle that has to be solved as quickly as possible, with many widely scattered, individual details.”
First Aid on Tracks
When vehicles have technical problems that train personnel can’t handle themselves, it’s a job for the service staff at Siemens Mobility Services in Erlangen. They evaluate the error profile and advise the vehicle driver as quickly as possible. In the future, Industrial Knowledge Graphs – a joint project of Corporate Technology and Siemens Mobility – will support service experts in their work. What previously took hours will now be possible within a matter of minutes.
A Service Puzzle
The puzzle pieces are the numerous items of information stored in various mutually incompatible systems: The vehicle number indicates which version of the complex on-board electronics is affected. Each individual vehicle also has its own service history that records what malfunctions and repairs the vehicle has already had. Other control data from the onboard electronics is also available, or there’s empirical data about past responses to similar problems. So there’s plenty of data that service experts can generally use to pinpoint errors, enabling them to decide what has to be done. But as past service calls have shown, even the experts sometimes require several hours to come up with a solution. Researching all the relevant information is extremely time-consuming.
From Hours to Minutes
This is where a joint project involving central Siemens research, Corporate Technology, and Siemens Mobility comes in: “Our goal is to make Siemens service faster,” explains Nils Weinert from Corporate Technology. “Based on artificial intelligence, we’re developing a solution that will provide service staff with a smart assistant – a sort of system that becomes a working partner, also known as a digital companion. It’s a tool that can independently recommend solutions and, in particular, takes over the tedious task of researching the various sources of data.” Whether it’s blueprints or the maintenance history of the specific vehicle or similar trains, the digital companion can access the information from the different data sources and knows how these facts are interconnected. Solutions that used to take hours can now be found in a matter of minutes.
A Knowledge Base composed of Industrial Knowledge Graphs
The basis for this solution is Industrial Knowledge Graphs, a mathematical approach that supplements artificial intelligence with an important capability: that of recognizing context and thus relationships. Taking the example of Mobility, these knowledge graphs can be displayed graphically as an extensive network. The individual units of knowledge – for example, a database for vehicle numbers or a vehicle’s service history – form nodes in the graph. The edges (connecting lines) indicate which nodes are related in terms of content. “This makes Industrial Knowledge Graphs ideal for recording semantic relationships,” says Nils Weinert – particularly since they can be supplemented with any new facts and relationships that arise, a characteristic that the current project is also exploiting.
The Mobility Knowledge Graph continues to grow
“The first prototype that we and colleagues from Service put together over the last fiscal year initially linked only order and service data to selected condition information of vehicles or trains. By linking the data sources, it was possible to find comparable service calls much more quickly and easily than before,” says Weinert. Over the next few months, the database will be supplemented with additional data from Service Engineering and Manufacturing. The basic plan is to fully integrate the Railigent diagnostic platform and the Rail Mall spare parts platform. In theory, the knowledge graphs will continually grow.
The knowledge graph project is embedded in a larger digitalization initiative by Mobility that is looking for solutions across Business Units. “With the knowledge graph technology, we’re moving one step closer to digitalization. Because we don’t waste any more time discovering information, we’re free to focus more on our customers,” says Thomas Müller. “People have already heard about how well it works. Ideas on how we can use Industrial Knowledge Graphs are being brought to us from other Business Units.