Researchers from areas as wide ranging as astrophysics, the automotive industry, and biomedicine use physical simulation and modelling software to aid the design and testing of products. The benefits include improved design, lower production costs and greater efficiency: Viasys Healthcare claims that the use of Star-CD from CD-adapco, a computational fluid dynamics (CFD) simulator, to redesign an infant ventilator device, reduced the design phase from one year to just three days and saved the company $250,000, while making it safer for the baby and cheaper to run.
This increase in usage has partly been driven by necessity: engineering processes are pumping out more data than ever before, and engineers require user-friendly software that can manipulate and analyse this data. In addition, computing power had always been a limiting factor, but the advent of cheap supercomputing has vastly increased the number of computations possible, which in turn has increased accuracy and speed, and allowed solutions to complex problems that previously would have been impossible.
Computational fluid dynamics is an area of engineering that has benefited most from this improvement in technology. The Navier-Stokes equations, which govern the flow of gases and liquids, can be impossible to solve exactly in certain circumstances. In these cases, a numerical solution given by a computer is the only possibility.
Like most simulation software, Star-CD uses a finite element method, where the fluid space being studied is split into discrete lumps (cells). It then performs a number of calculations for each cell, solving the Navier-Stokes equations, along with any other physical phenomena that may affect the flow. Stephen Ferguson, technical marketing manager of CD-adapco, says: ‘The increase in computational power has increased the possible number of cells used in calculations from 100,000 to 1,000,000. The cost has decreased dramatically, which has made it much more accessible.’ The user can choose how many cells he or she wants in the model depending on their computing power. In general, a greater number of cells give a better answer.
Typical of key development in recent years, Star-CD can now be integrated within the CAD environment of Star-CAD. It is now possible to run simulations while designing components within the CAD software, and then analyse the results and look for changes such as pressure drops. In this way, all of the major work is completed once, and then additional trials can be completed very quickly. This obviously increases efficiency significantly.
Another exciting development has been the ability to calculate thermal and flow stresses. ‘Sometimes people aren’t interested in the flows, just their influence on components,’ says Stephen. ‘Flows can cause structures to bend and deform. In addition, a temperature gradient from hot to cold can cause a thermal stress, which would be a problem in gas turbine blades in a jet engine.’ The integration of these within the CAD environment enables engineers to compensate for them in the initial stages of design, eliminating the need for excessive testing on expensive physical prototypes.
Stephen continues: ‘The number of applications is as long as a piece of string.’ As well as the baby ventilator, another high-profile application has been in the design of unmanned aerial vehicles, normally used for spying missions by the defence industry. The turnaround is much quicker than for normal planes, so analysis is needed much earlier in the process.
Daniele Camatti, CEO of ProS3 and designer of the Pro Observer UAV, says: ‘With Star-CCM+ we were able to estimate with great precision the position of the aircraft’s aerodynamic centre to such an extent that, from the first test flight onwards, we have completely avoided crashes.’ Star-CD is also used extensively within the automotive industry, in areas such as engine design and aerodynamic design. Formula One teams typically use it to create around five prototypes, which are then tested within the wind tunnels to find the best model. This validation is essential, as no model is 100 per cent accurate.
LMS and StatSoft both provide software that takes an even more direct role in the experimental process of engineering. LMS has two products that offer a greater integration of data capture and simulation. LMS Test.Lab includes a hardware measurement device with up to 300 sensors placed throughout the object being considered, measuring, for example, the noise level at certain RPM levels. The software is then used to find the relationship between the different phenomena, and find the root cause of the problem. LMS Virtual.Lab follows a similar process, but with the added option of providing virtual results for the tests, including motion simulation.
Although the pieces of software work independently, communication between them is possible as they share the same data formats. An example of this comes from the automotive industry. Up to 80 per cent of the components in the design of a car will come from existing vehicles, of which the acoustic behaviour will already be known, and can be imported into Virtual.Lab. Virtual.Lab will then incorporate the known data into its simulations, giving greater accuracy and avoiding time-consuming, redundant calculations. This recycling of data can reduce the analysis time from weeks to hours.
As Bruno Massa, director of corporate marketing for LMS, says: ‘This interchange of data is becoming an active part of design in automotive vehicles. Virtual simulation is very powerful, but when you put together the design of the whole vehicle it dramatically increases the size of the model. This creates an enormous burden on the computer power. The accumulated error might no longer be accurate enough to derive many conclusions.’
Increasingly, virtual testing is used to focus the time of testing on the least secure part of the design, and to find the weakest locations with the most problems from the very beginning. The possibility of more trials increases the probability of an optimum solution; by tracing the root cause of the problem it is possible to keep check on the side effects of these changes.
Applications range from the improvement of the performance of wind turbines, reducing unwanted vibrations in Heidelberg printing presses, and ensuring safe power switches in power stations, to analysing the sound profile of Harley-Davidson motorbikes to provide that genuine ‘Harley sound.’
However, applications of modelling and simulation software are not limited to engineering; Comsol and Wolfram have both developed software that are used for physics calculations. Wolfram’s Mathematica package is used widely in education. Originally built as a basic technical infrastructure for solving differential equations, it has since been extended and modified for more specific calculations. It automatically chooses the right algorithm for the type of differential equation.
Mathematica has a wide variety of applications, depending on the level of expertise of the user. Jon McLoone, sales and marketing manager at Wolfram, says: ‘At one end people use it is a super calculator; at the other end they use it to write complex simulation calculations.’ Apart from simulation, it has other built-in functions, such as the ability to perform calculus operations analytically, giving exact answers, and matrix manipulation, which set it apart from its competitors. It is these functions that allow scientists to use it for such physical problems as modelling of the hydrogen atom, tsunami prediction and the simulation of colliding galaxies.
Comsol Multiphysics uses a similar idea; the laws of physics are expressed as differential equations and entered via a graphical user interface. It has ready-made modelling interfaces for many areas of physics. For example, with an electromagnetism problem, the user does not need to write out Maxwell’s equations – she can simply give the physical properties such as the boundary conditions, and Multiphysics will do the rest of the calculations. Other areas with specific modules include waves, fluid dynamics, and structural mechanics. More recent additions include acoustics and MEMS.
It is a very generic mathematical engine, and can very easily couple any number of physical phenomena, which gives it far reaching applications.
Designers of generators have used the software to simulate vibrations and deflections, coupling electromagnetism and solid mechanics. Bioengineers also couple fluid dynamics and structural mechanics to model the valves in veins that prevent blood flowing back downstream.
There seem to be two common features to all of these products. Engineers and physicists have to work from many different areas, sometimes within the same project, and the integration of these into a single software package is vital to its success. In addition, physicists and engineers are not mathematicians, statisticians or programmers, so an easy-to-use interface is essential. Robert Eames believes that the biggest problem now is not processing speed or computational power – it’s acceptance by the scientists themselves: ‘We need to educate engineers about the benefits. They are oriented to practical, hands-on tools. The challenge is to show them practically that it does a better job.’
Data mining in engine design
Some products, like Statistica and its sister product Proceed from StatSoft, focus much more on the statistical analysis of test data. Proceed was developed as part of a collaboration with Caterpillar, the mechanical engineering company, to calculate the optimum parameters in engine design such as reliability, fuel usage, or power. It can do multiple simultaneous optimisations, and will optimise the parameters to a range of values, rather than a single value, which is difficult and costly to implement.
The statistical models used are often complex and non-linear. For this, Statistica uses data mining techniques including neural networks or recursive partitioning methods (tree methods) that are specifically geared to a large number of parameters. These powerful techniques could not have come at a better time for engineering. Robert Eames, from StatSoft, says: ‘We’re at a critical point in history; this is the convergence point from a number of processes. So many processes pump data out at an enormous rate and every manufacturer has huge databases. Over the past decade data mining algorithms have progressed until the modelling of this data was possible.’
Statistica's Proceed in action
DEM Solutions has developed a CAE simulation tool using a method called the discrete element method. The DEM method is more accurate for the simulation of discrete particles than continuum methods such as the finite element method, because it models each individual discrete object, and solves the problem locally rather than looking at the whole system.
The software can give information about each individual particle such as mass, temperature, velocity, and the forces acting on it. It can also take into account the particle’s shape, rather than assuming that all particles are spherical. Obviously this takes a lot of computing power, so scaling options are also available. On the post-processing side, data analysis tools and 3D visualisations of the particle flow are also possible, to analyse the interaction of the particle flow with the machinery.
Nasa is also using software from DEM Solutions to model the effects of electrostatic particles on their instruments. According to John Favier, CEO of DEM Solutions: ‘Nasa chose this software because it provides a very flexible tool to model electrostatic processes. Any physical phenomena can be coupled into the software, and used to model sophisticated electromagnetic algorithms.’ The software can be easily integrated with other software, and CAD designs can be imported to take account of the geometry of the instruments.
The software has been used to model erosion processes, conglomeration, sedimentation, particle damage and attrition. A typical pharmaceutical application would be in simulating tablet coating, powder grinding and mixing, and tablet compaction, to give information about how much damage the pills could receive through processing.
EDEM model of conveyor transport of particles
