Pictures of the Future    Fall 2007

Blending Realities

With eight factory halls, each as big as a soccer field and as high as a five-story building, the Siemens Electrical Drives Ltd. (SEDL) motor production facility in Tianjin, China (a two-hour drive from Beijing) is extremely imposing. Electric motors the size of a grown man are built here, as are wind turbines as big as small trucks, switching cabinets, and control units. Plans call for the Tianjin plant to be expanded even further and take its place as the leading facility for electric motor production in China.

But when the facility was originally built, it posed a major challenge because it had to be planned and built from the ground up within only two-and-a-half years. And, of course, you can’ t simply design a production location of such magnitude on the drawing board.

Because of the tremendous scope of the project, the Production Processes (PP) department at Siemens Corporate Technology (CT) in Munich was called in to help. The department specializes in creating three-dimensional factory computer models.

Long before the first bulldozer broke ground, components were moving along virtual assembly lines. The objective was clear: The more realistically a factory can be depicted in the planning phase, the more rapidly errors can be detected and avoided when actual construction begins.

Siemens specialists have been producing digital versions of factories for around 20 years now, and if they’ve learned one thing it’s that the best digital tools are useless if planners fail to understand factory processes in detail. "You first need to thoroughly review the entire planning process before you can begin using virtual tools," says Dr. Bernd Korves, head of the Production Networks & Factory Planning competence center at CT PP. The key here is to completely understand the entire lifecycle, from design all the way to suppliers and production. Experts refer to this as product lifecycle management (PLM).

The result of the design process—a digital product—is the bridge to the digital factory (see Product Development). "Extensive interlinking of these two process blocks offers huge potential," says Dr. Albert Gilg, head of the Virtual Design competence center. "That’s because product design ultimately determines whether you create obstacles to production or enhance the efficiency of the manufacturing process."

Design data is thus the point of departure for the extensive analysis of a future production system. Specialists determine which production steps will be necessary and the optimal sequence and speed of those steps. They determine the kinds of workstations needed for each step and how a factory should be laid out. Planners then work with the relevant Siemens Groups, often coming up with several alternatives. Each proposal is depicted on a computer as a 3D factory whose operations, including material flows, is simulated in detail. Throughout this process, planners make extensive use of digital libraries to visualize individual workstations, machines and processes.

With the Tianjin plant, virtual models enable planning teams around the world to "fly into" the factory halls at the push of a button. Large gray tubes can be seen in the halls—the motor stators. Next to them are avatars—simulated humans who grab copper wires and insert them into the tubes. The virtual flight allows employees at SEDL to quickly identify whether each workstation has enough space to move large motors around in, for example. Changes can be incorporated at any time, and their impact is immediately shown in the simulation. One special challenge in the Tianjin project was the fact that the virtual facility went through a simulated development of more than five years, which means production capacities had to be expanded as time went by and changing demand for products had to be taken into consideration.

Degrees of Abstraction. The art of simulation mainly involves being able to figure out which locations require detailed information from the real world. "A lot of beginners try to precisely reproduce reality, which is a mistake," says Korves. It’s also counterproductive because it requires way too much effort and expense. Success here depends on determining the proper degree of abstraction. "If you’re simulating material flows to come up with a layout, you don’t need to have everything down to the smallest bolt—but you do need this kind of information for complex assembly simulations," Korves explains.

Korves did in fact have to get very detailed in another project he worked on with Siemens VDO that involved production of a new vehicle dashboard. The job required detailed depictions of manufacturing cells as a means of simulating their ergonomic properties. Here, CT used software from UGS, which is now part of Automation and Drives (A&D) and is known as Siemens PLM Software (A&D PL—see UGS and Siemens). The software utilized standard values to record the size and stature of a worker and the number of times he or she repeated certain movements. This made it possible to optimize workstations by adjusting things like bench heights and arm-length distances to neighboring machines.

Most virtual models today are created using objects from digital libraries. The CT team’s expertise lies in its ability to come up with the best solution for each application, even employing its own user interfaces in some cases. For instance, in cooperation with Munich Technical University, the team came up with Plant Calc, a sophisticated planning tool. Plant Calc software can compare production locations using a systematic assessment of various alternatives, which also takes into account planning uncertainties. In a study conducted by CT for a Siemens plant in northern Germany, Plant Calc determined that under certain conditions, expanding production in Germany would be better than transferring it to Eastern Europe. The study found that although wage costs in Germany are higher, the potential for optimization in the country made it a more economical production location.

True-to-life Virtual Testing. Reality and the virtual world are moving closer together at A&D, which operates two "SmartAutomation" research centers in Nuremberg and Karlsruhe that will be used to develop automation solutions virtually and in real life. Researchers have set up a bottling facility in Nuremberg and a chemical processing unit in Karlsruhe, both of which enable new ideas to be rapidly implemented in actual equipment for the first time. Among other things, researchers are now building a robot that grabs bottles as they pass by, takes them to a quality control station, exmines them, and returns them to exactly the right spot on the production line.

All of this was planned and tested in the virtual world. To do so, A&D developers inserted the virtual robot into its future real position in an image of the existing facility. All bolts, measurements, electrical connections, data communication and pressure systems were verified before actual implementation. The researchers even ran a realtime simulation of the robot’s operating parameters. On the other hand, the initial data entered into the system for simulating the bottle-picking robot came from the physical bottling unit. "The fascinating thing about SmartAutomation is that you can directly link reality and a simulation," says project manager Bernd Opgenoorth.

Despite the excellent performance of the simulation system, there is still room for improvement, especially with regard to the comprehensiveness of the planning process. That’s because data from the entire process chain does not pass seamlessly from the first draft design to the finished factory model. In many cases, data has to be transferred manually from one program to the next—for example, from a 3D drawing to the visualization software, or from a virtual model to the language used by a computer controlled CNC milling machine.

"What we need to do now is eliminate the discontinuities and automate the transfer of data from the beginning to the end of the process," says Opgenoorth. Researchers from his team are working with A&D PL to solve this.

Lego for Factories. A similar approach is employed by the "SmartFactoryKL" project managed by the German Research Center for Artificial Intelligence (DFKI) in Saarbrücken. The center is a consortium of companies and research institutes that is also working on a miniaturized version of a real production facility. A founding member of the consortium, Siemens A&D also provides funding for the SmartFactory, which, like SmartAutomation, simulates production in the virtual world. One of the factory’s purposes is to demonstrate how components from different manufacturers can be combined. It’s a visionary idea that foresees having factories built from standard modules much like giant Lego blocks. This would require that each producer’s modules be equipped with standard interfaces.

In addition, all SmartFactory plant components for the miniaturized production facility are to be equipped with radio frequency identification labels (see Factory Data Democracy), thereby making it possible to automate inventory registration and precisely pinpoint machine locations. This, in turn, will make it easier to expand or convert existing factories. Machine locations could be fed into virtual models to enable planners to determine exactly where new equipment should be installed. "A lot of work—and information—goes into virtual factory models," says DFKI project coordinator Eric Pohlmann. "So it makes sense to use this great variety of data over and over again."
                                                                                                                  Tim Schröder