High-speed and commuter trains are extremely complex constructions with thousands of components. During the validation and homologation phase, actual trains are rare and expensive assets. Siemens Mobility Rail Solutions sought to develop a cost-efficient way to test high-speed and commuter trains.
Siemens used NI hardware, TestStand software, VeriStand software, and the LabVIEW FPGA Module to build a fully functional digital twin of a whole train, including a functional simulation of all the wiring. Train control, traction, and brake controllers are integrated as real devices, but the system is easily extendable for any other controller.
Matthias Reinholdt - Siemens AG Mobility Division
Jerzy Kocerka - Tritem Microsystems GmbH
In 1879, Werner von Siemens invented the first electric locomotive. Today, Siemens Mobility Rail Solutions focuses on rail infrastructure and rolling stock. As a part of the rolling stock segment, we develop solutions for high-speed and commuter rail customers and are a global player in 60 countries around the world.
The rail and transportation industry has grown since the industrial revolution, with many regional differences. Each train is a customer-specific solution that must be integrated in an operator-specific environment. For regulated homologation processes, many requirements must be validated and approved. The train itself is a rare and expensive asset, so we need to find solutions that address the validation and homologation requirements with a more cost-efficient test environment.
There have been different train simulation solutions in the past, but they usually implemented a behavior of the interfaces to test a dedicated single controller. This application realizes the complete logic behind the interfaces and creates a realistic behavior of the train, including its whole electrical construction. Compared to earlier solutions with fixed dedicated test environments, this application is a highly scalable solution. An iron train is a physical, mocked-up train that emulates all the inputs and outputs of a real train. We could run the digital twin of a train on a laptop as well as at the level of an iron train, which makes this application the first of its kind. We designed it for use with many different variations of trains requested by our customers. This strategic investment is the backbone of all future Siemens Mobility projects in the high-speed and commuter rail market.
In 2011, the German rail company Deutsche Bahn AG ordered 170 high-speed trains with an option of up to 300 trains for more than $7.4 billion from Siemens Mobility. These trains, known as ICx, are completely different than conventional constructions, which reflect the changing demands of our markets. We based the architecture on a power car concept and a flexible train configuration.
This concept should address high flexibility in terms of train length and passenger count. We invented a new car-based train control architecture. Each single car of the train has an encapsulated train control system. Every new architecture introduces changes such as new communication bus technology, new automation systems, and new interfaces. To address the changes and the corresponding risks of new technology, we need adequate quality assurance.
Using PXI real-time devices, NI EtherCAT chassis, VeriStand, and a variety of I/O modules, we created a test bench that simulates complete train functionality of 40 different subsystems. We took advantage of the NI product portfolio for a complete set of I/O interfaces, a stable run time, and tooling setup, which is difficult in many industrial applications. There was no other solution that better suited our various interfaces and requirements in terms of modularity.
The main feature of our solution is the functional simulation. The core of this functional simulation is the representation of the electrical schema of the train in the simulation environment, which can be visualized and manipulated in real time. We import this electrical schema into VeriStand as an electro-mechanical logic model, which, combined with models that represent the functional logic of the physics and complex elements, like simulated controllers (for example, a door controller), enables us to build a complete digital twin of a train. Overall, we created a model library with 350 intelligent elements, like controller models, and 150 electrical models, like connectors, switches, and relays. Out of this library, a VeriStand project generator automatically creates a complex hardware-in-the-loop (HIL) system with more than 58,000 simulated components.
Finally, the test bench consists of 16 test racks: 12 cars of the train and four additional racks with test bench functionality such as fault injections. Each car test rack includes a complete train control system, the communication buses, two brake controllers, and, in the power car types, the traction controller. We could integrate real components like this because the high reliability and determinism of well-integrated FPGA technology helped us simulate all necessary I/O interfaces (for example, 192 speed sensors). Each single rack is like a complete HIL environment.
We realized the simulation hardware with 12 PXI systems and and 42 NI-9144 EtherCAT chassis. We use PXI-6683H timing and synchronization modules to synchronize the PXI systems. We also use several different analog and digital CompactRIO and PXIe-2727 programmable resistor modules as interfaces to the train controller devices. Our test bench implements CAN bus, PROFINET bus, and other TCP-based protocols.
We could automatically generate nearly 70 percent of the model logic and VeriStand mapping from engineering and construction data sets. We imported the electrical schema data set out of an ECAD system. We generated the bus interfaces from digital interface specifications of the individual communication participants. This leads to a 100X cost reduction compared to an iron train.
With the simulation environment, we can test every controller of a modern train. We can exchange a simulated component with a real one just by changing the mapping in VeriStand. For better development and usage, we created several add-ons to the NI ecosystem.
Besides UI panels for manual testing, we can visualize the whole electrical schema and the electrical states in real time. We can also manipulate the simulation using this visualization. For example, we can cut wires, change electrical potentials, or freeze relays.
We can also use test automation combined with a test management system. The test cases are created in the test management system and a code generator generates TestStand sequences. We can synchronize the results with the test management.
We can transfer many test activities to the laboratory to save costs and resources in customer projects. The ability to force scenarios that are expensive or impossible with a real train reduces the risk of undiscovered failures and impacts to the project timeline dramatically, and by simulating the train with such detail, we could inject faults at almost any position, which allowed us to create test cases that are very difficult to set up in reality. The massive number of signals and variables are calculated in a 48 ms loop rate over the 12 PXI systems, which are synched with 500 ns accuracy. Another benefit was the reduction of travel expenses since we could verify many bug fixes and feature improvements in the laboratory.
The significant impact of the evolutionary step in quality assurance is to have a complex, realistic replica much earlier in the development process to reduce non-conformance costs and risks. This simulation application supports the development, integration, test, validation, and homologation processes of controller components. The high amount of generated simulation content enables us to react quickly to changes and ensure a high quality of the simulation application itself.
Testing a train is expensive. Testing with infrastructure efforts can cost more than $10,000 per day. Considering non-conformity costs and project milestone delays in the past, using this simulation application reduces cost by $1 million to $8 million per project. Additionally, after importing all the necessary data, we can now create a new train variant and start testing in less than two hours.
The scaling capabilities of the NI ecosystem and our simulation architecture helps us reduce the number of hardware components. This results in a reduction of physical interfaces that need to be simulated. These are the cost drivers of building the test bench.
Our goal is to provide a train to everybody, whether it is as a tester, developer, supplier, or customer. We accomplish this through simulation, regardless of whether the simulation is used for testing, demonstration, or as living specification for subcontractors. Providing a train to everybody could increase the quality of supplied components and development output. Training for customers could be started earlier and in a dramatically cheaper environment than the train itself. This kind of application will be a game changer in our business.
Siemens AG Mobility Division
91052 Erlangen, Germany
Tel: +49 173 7312116