Two of the biggest issues facing appliance design engineers today are cost reduction and customer satisfaction. Excessive defects in the development of appliances and variation in materials and manufacturing processes generate unexpectedly high costs. Additionally, consumers are not content with yesterday’s performance and style. They demand better and quieter performance and more style. At the same time, company initiatives allow less time and cost for developing these innovative, yet lower-cost products. Enterprise multidiscipline simulation solutions enable collaboration while allowing far earlier use of simulation to create category killer appliances that are durable, quiet, meet safety standards, cost less to produce, and dramatically reduce time to market.

Single-point tools

Appliance manufacturers have already implemented single-point simulation tools to aid product development. The most common use is forensics — trying to figure out what went wrong when a physical prototype fails. However, single-point solutions do not consider the interaction between various disciplines that can cause failure. Even when bundled together, single-point simulation tools cannot provide as accurate a solution as a tightly coupled, multidiscipline solution for strongly coupled problems.

Coupling disciplines

Fig. 1. Cut-away view of virtual clothes-washer basket and suspension.

Only an accurate representation of the complex interactions between key disciplines can ensure simulation of the physical phenomena that approaches real life. Coupling such disciplines as thermo-structural, non-linear dynamic, fluid structural interaction, acoustics and electromagnetic structural, results in predictions that are more realistic. Yet, even with recent advances in modelers (pre and post processors), and advances in computing power and automated capabilities, simulation discipline specialists still manually simulate the complex inter-discipline interactions as discrete analysis steps. Assessing large volumes of analysis data to determine how to hand off results from one discipline to another is inevitably many times more tedious, more subject to human error, more compromising to simulation accuracy, and often unrepeatable.

Sometimes an engineer carries information by hand or forces the information from motion in a static manner to impact the FE representation of a system. By connecting them, the information becomes live, and they are in an open loop environment. Whether it is linear, nonlinear, motion, CFD or explicit dynamics, this allows disciplines to work together, rather than simply communicate with each other.

Working together implies the disciplines provide correct engineering and mechanical feedback to each other at exactly the right time. Moving beyond traditional multi-physics systems, discipline chaining/integration between multi-body motion and FEA facilitates simulation capabilities allowing enterprise-wide multidiscipline simulation to drive design early in a product cycle, such as external system loads spectrum definition. The same is true with integrated FEA and CFD analysis.

Common Model

A system level model, a single data model with multiphysics representations, allows engineers throughout an enterprise to work from the same data.

An enormous benefit of coupled multidiscipline simulation is the use of a common analysis model, which drastically reduces design errors and analysis cycle-time that impacts product development decisions. By eliminating separate models for each point solution, the time spent generating, maintaining, storing and searching for them is eliminated. Based on a system level model — single-data model with multi-physics representations — engineers throughout an enterprise and the supply chain can work from the same data. When compared with bundled single-point simulation tools, time to solution reductions of up to 50 percent can be achieved.

A common model does not infer use of a single model across every discipline, but that extractions from a common model are used to create representations of systems for simulation with common loads and constraints. Not every simulation problem is solved with a single equation. It requires a series of equations working with common data to provide the most realistic simulation possible. A finite element model is an abstraction of a physical system that can be loaded and conditioned to study various disciplines.

The disciplines that must interact with each other guide and determine whether analysis is strongly, staggered, or loosely coupled. Strongly coupled environments require that the disciplines interact and the correct results are heavily dependent on these interactions. Staggered analysis means a series of analyses has to be performed in order to produce correct results. On the other hand, loosely coupled problems mean that the disciplines act independently of one another and considering the interactions does not affect the accuracy of the results. A system level model combining motion studies with FEA and CFD offers many applications for multiphysics. For example, motion with a load of clothes in the washing machine. The masses (clothes and water) are continuously moving and changing the required forces to operate.

Optimization

Users can run optimization loops at various levels of simulation. Shape and topological optimization operate within each discipline, and variability (stochastic or probabilistic) optimization determines the robustness of the design. Engineers can look at system equations and determine variability at all levels — especially where material properties and manufacturing processes have a great deal of variability.

Stochastic optimization is known as randomization; it targets thicknesses, material properties, spring stiffness, beam cross-sections, forces and imposed displacements, etc. Each tolerance is defined by engineering limits and a distribution function. The model is then executed a certain number of times (50 to 100 is typical), randomly changing each of the input variables within assigned tolerances. The more tolerances, the more realistic it becomes. If a model has thousands of variables, then all should be randomized, not just the ones that are thought to be decisive and influential.

The more complex a product is, the higher the chance that some unpredictable combination of uncertainties will have an unexpected and potentially undesired effect. This phenomenon manifests itself in the form of outliers, or uncharacteristic behavior, which can be translated into risk and liability. Anticipating their presence allows an understanding of the circumstances under which they arise.

The primary result of a stochastic simulation is the most likely behavior of the system. Since the most likely performance does not coincide with the nominal one, this is extremely important. Because the distribution of performance is obtained, an engineer may be able to determine if manufacturing quality is an issue. Stochastic simulation provides an opportunity to not only consolidate this knowledge, but also to extract additional, often unexpected information about a system.

High-performance computing

Common analysis models coupled with multidiscipline simulation reduces design errors and analysis cycle time.

More sophisticated and complex simulation models with unbounded model and analysis data-set size have increasingly become a necessity. At the other end of the lifecycle, engineers simulate entire systems prior to manufacturing. Even before considering discipline integration, it is clear that model size and complexity is continually growing. The ambitious scope of enterprise simultaneous multidiscipline solutions means they must handle very large problems. And, to make very large simulations fit within today’s time constraints, programs must run on both 32-bit and 64-bit computer cluster environments.

When interdiscipline simulations are factored in, an even greater burden is placed on computational resources and demands. These models can be overwhelming in size. However, because of new technologies such as 64-bit processors and SMP/DMP (shared memory parallel/distributed memory parallel), which allows multiple computers to work the same problem simultaneously, these very large problems can be solved quickly. Additionally, techniques such as superelements divide the model into groups that are much faster to process alone than separately without affecting the accuracy of results.

Conclusion

To continue to satisfy the high demands of their consumers, appliance manufacturers must perform interoperable multi-disciplinary analyses on system models (parts and assemblies). Enterprise-wide coupled multidiscipline coverage addresses a broad set of true multidiscipline problems with more accurate and reliable performance predictions. This is the power that puts customers one step closer to the virtual product development environment of the future and contributes to the development of category-killer appliances.


For more information email: carlson.choi@mscsoftware.com