Pictures of the Future    Spring 2006

Paths to Perfection

Less than two decades from now we will be able to see, hear, and interact with hyper-realistic products, processes, materials and environments in their full dynamic diversity long before they exist. Simulation and optimization will make it all possible.

Advanced simulation systems are set to usher in an era of virtual prototypes, virtual testing, virtual training, and knowledge-guided optimization that will eliminate errors, cut costs and vastly accelerate the development of everything from engine parts to entire factories.

How close are we to this vision of the future? In some cases, such as virtual training (see Training), the basic technology is already well established. Other applications, such as the use of existing software on virtual prototypes—a new dimension in simulation that promises to accelerate development of particularly complex systems—are just beginning to enter the mainstream. Two to three years away, depending on the complexity of the application and on advances in computing speed, we will see the introduction of so-called "mixed simulation"—the combination of different parameters, such as acoustics, flow dynamics and function, possibly in real time. And between five and ten years from now, experts foresee the first applications of "multi-scale modeling"—the melding of nano level data with entire virtual objects, thus nudging simulations closer to reality while bridging the gap between two categories of simulation that are worlds apart today. Along the way, new algorithms, some based on neural networks, will learn to harness the immense information content released by the simulation of previously invisible worlds, thereby opening novel ways of optimizing the objects and systems we simulate and build. Eventually, by 2015 or so, all of these trends are expected to converge. "At that point," says Dr. Wolfgang Rossner, head of the Ceramic Materials department at Siemens Corporate Technology, "we will be able to develop virtual engineering prototypes based on everything from the nano to the functional level in a virtual real environment—and have it accurate."

Where are we today? The articles that follow provide impressive examples. In China, where Siemens plans to build a production facility for small electric motors, expensive planning errors will be avoided thanks to a full scale, 3D simulation of the entire facility, complete with material flows and control electronics. (see Virtual Reality Lab). And thanks to simulations now being conducted by Siemens, the eight million residents of Brooklyn and Queens will benefit from faster commissioning of a planned direct current line that will deliver an additional 750 MW of power to Long Island by 2008 (see Testing with Simulation).

Formulas for Efficiency. Although simulation and optimization often go hand in hand, optimization can be applied as a stand-alone solution in many cases. For instance, optimization algorithms have improved the through-put of assembly lines in the microelectronics components industry by 13 % and have considerably improved the efficiency of steel production lines (see Optimization by the Numbers).

Postal authorities and parcel delivery companies also stand to benefit from optimization algorithms. For instance, thanks to a device called an Intelligent Singulator, which was developed at Siemens Corporate Research (SCR) in Princeton, New Jersey in collaboration with Siemens Dematic and the Siemens Technology-to-Business Center in Berkeley, California, a bulk stream of parcels can be separated into a single file for sorting. The Singulator uses a multi-camera vision and control system to provide parcel locations and orientation in real-time to a motion control optimization algorithm that controls the speeds of conveyor belts to rotate and separate parcels into single file. "The system has been operational at UPS Cologne since January 2006 and has already proven that it improves collision-free parcel extraction and eliminates carton wear-and-tear, while significantly reducing noise levels," says Yakup Genc, who heads SCR’s 3D Vision and Augmented Reality program in its Real Time Vision and Modeling Department. Thanks to Siemens technology, the Cologne facility can sort some 110,000 packages per hour.

Unmatched Precision. Optimization algorithms are also set to improve the efficiency of medical imaging systems. At SCR, for instance, Leo Grady, Ph.D., has developed and patented a groundbreaking algorithm that will soon make it possible to quantify tumor size with a previously unheard of combination of speed and precision. After a radiologist identifies one spot on a medical image as being part of a tumor (object one) and other spots as being background (object two), the algorithm, called "Random Walker," defines the probability that each pixel will belong to one or the other of these regions by defining an affinity between pixels based on established biases. Essentially a segmentation optimization question, the key, explains Grady, "is to identify some relevant quantity, and minimize or maximize it. The algorithm then distinguishes between tumor and non-tumor areas by weighing the affinities of each pixel." Siemens Oncology Care Services expects to deploy the algorithm in its COHERENCE radiation planning package to help physicians precisely see the 3D contours of tumors as well as those of sensitive areas, such as the eyes, that must be avoided. In addition, the product holds the broader potential of providing valuable feedback for patients by measuring the extent to which a tumor has changed in size between tests.

But optimization may have a broader field of applications. Looking to the future, Prof. Martin Greiner, a theoretical physicist with Siemens Corporate Technology, says: "My vision is a form of distributed optimization in which complex networks function as if they were intelligent societies." An example is an invention Greiner recently patented that will allow the rotors in wind parks to optimize the positions of their blades in response to information provided by the first towers affected by a change in wind speed or direction "This will help to reduce power fluctuations and will allow an entire wind park to function as if it were a single entity," says Greiner.

Such networks—whether made up of windmills, motors or molecules—will learn from experience and will become interconnected in a virtually biological and social alliance. "Understanding these networks, their dynamics and architecture will be a major focus of learning systems and neural networks for the next ten to twenty years," predicts Prof. Bernd Schürmann, head of Siemens Corporate Technology’s Learning Systems department. "Forms of cooperation take place in cells, among ants and wasps that will have implications for the Internet, for traffic management, for logistics and much more. As we learn to extrapolate knowledge from natural systems, we will use it to optimize human systems and society itself" (for more, see Learning Systems).

Engines of Tomorrow. Like the knowledge revolutions triggered by the development of the telescope and microscope, simulation and optimization hold the potential to open fundamentally new avenues of human exploration. "Simulation and optimization will enable us to explore areas beyond any that were accessible before," says Prof. Albert Gilg, head of the Simulation and Risk Management department at Siemens Corporate Technology. Gilg points out that the first step these fields are making possible is virtual engineering and the transition from physical experiments to virtual prototypes. "But looking ahead," he says, "they will enable us to work with computers in new ways, and to detect new principles, including things we have not even thought of."

To get to that point, says Gilg, it will be necessary to develop new classes of algorithms that can take today’s fundamentally separate fields of mathematics into account. These fields are:

Efforts to combine these disciplines are bearing fruit. For instance, until recently, estimating the maximum and minimum efficiency values of a proposed turbine blade design for a jet engine required running "months of simulations," says Gilg. But a new class of patented algorithms developed at Siemens Corporate Technology that combine ideas from stochastics with numerical algorithms "reach the same results with less than a tenth of the effort of previous methods," he says. "It’s a real breakthrough because the system, which we call RoDeO (Robust Design Optimization), not only works much faster, but delivers a more accurate representation of reality. As a result, for example, we can now guarantee with accuracy of at least 99 % that an engine will deliver a value within a predetermined range." What’s more, he adds, "In bridging two of the key fields of mathematics, we’ve solved a set of problems much more efficiently than ever before. And these new algorithms have a huge variety of potential robust design applications."

Sky’s the Limit? While new algorithms are turbo charging the visualization of gigantic industrial systems, others are helping scientists to peer inside nanoscopic worlds. For instance, in Dr. Wolfgang Rossner’s laboratories, researchers are investigating how to build a virtual object from the ground up. "One of our goals is to begin with virtual materials, define their microstructures, shape them, assemble them into a device, and simulate the device’s behavior," say Rossner.

Although researchers are still a decade away from such ready-to-use advanced multi-scale modeling technologies, Rossner’s group is already developing simulation tools for ceramic materials that will combine the atomistic level with the micro structural level. Such a convergence of capabilities will be crucial for the development of ceramic coatings capable of maximizing the extent to which industrial gas turbine blades are thermally insulated from hot gasses. "While detailed modeling of thermo-mechanical loading and degradation mechanisms has already led to a significant extension of operating temperatures of turbine blades—resulting in turbine efficiency increases of several percent—the next generation of thermal insulation will need ceramics that optimize the atomic composition and specific nano-scaled micro structural characteristics of these coatings. This goal can be achieved only through the intensive application of multi-scale modeling," says Rossner.

Of course, much "bigger" adventures are in store. "In the near future we will combine parameters such as acoustics, magnetic fields, flow dynamics and function," says Heinz-Simon Keil, head of Virtual Engineering at Corporate Technology. "I expect that within two or three years it will be possible to simulate a six-cylinder engine in real time, complete with sound and flow dynamics." And, as simulations of production environments are bound to keep pace, the parallel engineering of products and their related production lines—including associated software—will become a routine matter of clicking data sets back and forth between groups of engineers.

Simulation and optimization. These two words will shape the industrial landscape of the 21st Century. "In just fifteen years it will be possible to design a new car, test it, optimize it, and send the simulation to production without ever building a physical prototype," suggests Keil. Sure. Perhaps the sky is the limit. But there are still skeptics. "I think there are some things we’ll never be able to simulate," cautions Dr. Johannes Nierwetberg, who heads the Discrete Optimization department at Siemens Corporate Technology. "For instance, how do you write a program that can accurately predict the weather at your home several weeks from now?"
                                                                                                                Arthur F. Pease