vol 9, issue 3

Modelling and simulation innovation: just do the math

14 March 2014
By Martin Courtney
Share |
COMSOL’s Multiphysics

Mixing it: COMSOL’s Multiphysics depicts heat transfer from the pipes/walls of a modelled mixer

The Shredder personal off-road vehicle

Dassault Systèmes SolidWorks simulation tools were used to design and stress test The Shredder

QBall-X4 quadroctor helicopter

Design engineers at Quanser used Maplesoft’s MapleSim to develop the QBall-X4, a quadroctor helicopter

Rolls-Royce backs BLOODHOUND, credit Stefan Marjoram

MathWorks MATLAB was used to build a simplified model able to predict the performance of the Bloodhound supersonic car

SKYACTIV-D engine

Mazda used MathWorks model-based calibration software to improve their SKYACTIV-D clean diesel engines

IntelliFill agricultural equipment

MathWorks collaborated with agricultural vehicle manufacturer CNH to build IntelliFill, an automatic trailer filler system

Fidesys Online web service

The Fidesys Online web service allows to carry out deformation mode analysis for articles under static loads

Modelling and simulation tools are increasingly using powerful maths and physics computations to engineer a wider range of products faster and more accurately.

The latest generation of modelling and simulation (M&S) software is being used to supplement or supersede more traditional computer-aided design (CAD) and manufacturing (CAM) applications, enabling engineers across a range of industries to more rapidly design new projects, ranging from smart vehicles to fashion accessories. The more advanced systems are also increasingly able to conduct virtual prototyping and performance testing, and see where the proposed model has unintended or undesirable effects before the final product gets anywhere near the physical manufacturing process.

More powerful high-performance computing (HPC) platforms are helping to facilitate the advances. High-end workstations, servers, and supercomputers are being tasked with number-crunching complex maths and engineering calculations. These use a much broader range, and higher volume, of data sets to meet the initial requirements of design processes when 'developers are concerned less about the details of the final composition than the constraints and rough dimensioning of the model under development.

Another important factor is M&S software's ability to repurpose elements that were initially developed for previous projects. "We pick up where standard CAD/CAM software stops, primarily looking at system dynamics rather than physical appearance, and addressing structure rather than the form of metal or plastic," says Jos Martin, principal software engineer at technical computing software developer MathWorks. "We import from that list of bodies or parts - and how they connect together - [then] model how they move, add motors, hydraulics, different physical domains and simulate those to describe how they would move, how much power they would require."

If you have failures in the prototypes and tooling issues, you can "get the right product out of the door much quicker", adds Jonah Normand, product specialist in design, lifecycle and simulation at 3D design software firm Autodesk. "There is a lot of research around fibre- and carbon-filled materials in, for instance, making vehicles more lightweight. The manufacturer can experiment with different materials and have the confidence to input different figures to analyse the stresses on material and fibre orientation, and take that stress analysis to get a much more accurate result."

M&S for smaller enterprises

The way in which advances in modelling and simulation are influencing product design and conceptualisation are not immediately obvious to the end-consumer, but are having profound effects on many industries. Some industries may not have been able to afford using such sophisticated tools before, and are now enabled to do so by emerging factors such as the M&S software companies desire to broaden their prospective clientbases, and the availability of HPC resources on a pay-as-you-go basis.

The comparative glut of raw processing muscle now available has extended the reach of M&S software into smaller enterprises than previously, but indications are that even larger manufacturers with mature computer-aided engineering environments are being forced to optimise the design process due to cost and time constraints.

"It is clear that the complexity of the designs being worked on has grown tremendously over the last five years," believes Dr Laurent Bernardin, executive vice president and chief scientist at M&S software tools provider Maplesoft. "The constraints on control systems and fuel emissions are tighter across many different industries for example, and the scope for time saving and making changes for errors has increased by an order of magnitude."

The latest version of COMSOL's Multiphysics software suite, released last November, introduced several enhancements to the software's existing functionalities which exemplify the application areas M&S is now required to enter. The Acoustics Module, for example, offers aeroacoustics simulations based on the linearised Euler equations: these aim at enabling better simulations of noise in jet engines, mufflers, and gas flow meters, for example. Surface wall 'roughness' for turbulent flow, has been added to the Multiphysics Computational Fluid Dynamics Module.

The AC/DC Module for electrical projects now contains a non-linear magnetic material library. A new feature in the RF Module enables the simulation of components with ports on interior boundaries. The Wave Optics Module now includes scattering with a Gaussian background field and Laser Heating interface. The Semiconductor Module has also been upgraded to include heterojunctions and tools for impact ionisation.

Diverse datasets increase accuracy

One of the reasons behind the improved featuresets and accuracy of modelling software lies in the wider range and increased volume of information which can be input into individual simulations, ranging from energy consumption to climate statistics, in an attempt to gain a better idea of how models 'behave' using different materials in varied 'scenarios'. The data sets are a mixture of information used in similar scenarios, or new numbers which are adapted for specific applications - a new manufacturing facility based overseas, say, might have a different climate, but will re-use site measurement and temperature data, for example, while new cars based on traditional petrol engines will use existing data sets.

"Lots of models are developed from previous machines - it is very rare to design a new car from scratch, for example," explains MathWorks's Jos Martin. But there are exceptions: hybrid vehicles are new and there is very little stuff they can re-use. Martin adds: "So there you start from the basic physics, parameterise the data sets, test them, and evolve the modelling until you have the final outcome."

Being able to input and process so much information to obtain the best results is due in no small part to the increase in hardware compute capacity now available to engineers and developers. On-premise servers and supercomputers with much larger complements of CPU power, memory and storage allow for more compute-intensive modelling and simulation batch jobs to be run in shorter timescales.

Maplesoft offers multi-core processing capabilities contained in the clusters of workstations and supercomputers, which have become readily available to large corporates and other organisations.

The company has written algorithms specifically to take advantage of larger complements of processing capabilities, and also of the fact that some users are writing their own scripts to take advantage of multi-threaded computations. "That is handled automatically as much as possible and no user intervention is required, though users can go into full control to coordinate and synchronise jobs themselves," says Maplesoft's Bernardin, who cites one customer that ran a simulation of theoretical physics on a supercomputer based on over 100,000 CPU cores. "That [automation] is important because doing this manually and exercising full control often demands a level of sophistication that most people are not prepared to learn about," he adds.

Cloud allied to HPC

MathWorks is also conducting trials into how large HPC clusters based in hosted, off-premise cloud environments can help its clients. This includes the use of public infrastructure as a service (IaaS) platforms offered by Amazon Web Services and others.

However, the company says that some end-users are still uncomfortable about trusting mission-critical data and applications to third-party data centres, voicing concerns over security, performance, reliability, and a loss of control. After all, CAD/CAM data often constitutes intellectual property of the most sensitive nature - and highest value.

"We are finding that larger companies are more worried about it than smaller ones, which are generally more happy to try new concepts around building [HPC] clusters on demand," reports MathWorks' Martin. "We enable the cloud as just another scheduler, so we are seeing a bit of a broadening of the market - but the whole thrust of the HPC initiative is that traditional clusters are hard to use, and specific to the solution in question, which means you tend to be tied into one particular customer and vendor. With the cloud, the customer gets to choose where they want to run their jobs."

This suggests that although a customer may be outsourcing provisioning and management of the underpinning compute resources, they might still need to have the competences and understanding to properly specify their application needs. But now help is at hand for even that side of things.

MathWorks has integrated support for public cloud environments into its platform, allowing customers to 'cloud burst' demanding jobs into instances of MATLAB hosted on Amazon Web Service's Simple Storage Service (S3). "Cloud providers have come to us and asked if they can host MATLAB and the attraction with AWS is that it offers the ability to control what is going on," Martin says. "Others [cloud service providers] are starting to catch-up, but they are not as widely used as S3, and so it made sense to collaborate with the provider offering the best data footprint in Europe, China, the US, India, and Australia."

There are other examples of this model. One of the teams participating in the UberCloud HPC Experiment - BioTeam - also harnessed AWS's Elastic Cloud Compute (EC2) IaaS platform for virtual prototyping and design optimisation of sensors and antenna systems for use in medical imaging devices in addition to its own, local HPC clusters for availability purposes.

Product lifecycle management

Keeping track of complex, lengthy M&S projects involving multiple contributors and diverse key performance indicators (KPIs) is challenging to all involved. This is why such applications are increasingly integrated with product lifecycle management (PLM) tools, an approach which allows data to be shared more easily without having to import/export that information to/from different applications. This is particularly useful in manufacturing to ensure objectives are set and met with elements of part planning, shop documentation, resource and and data management, and data exchange.

"They start with a model for requirements, generation and capture, and allow you to play with ideas and concepts and work out how performance targets will be met," says Martin. "That evolves throughout the project with the model as its central point, including parameters and features, production code for the controller and documentation for safety."

ANSYS SCADE LifeCycle automatically also generates appropriate documentation and tracing elements to assist in this process, while Wolfram's Mathematica' product is a workflow application which keeps track of computing results in a variety of scenarios, helping to move ideas to deployed products. Similar project management tools are built into Wolfram's Workbench development platform and SystemModeler CAD/CAM application.

Aerospace exemplars

Two industries - aerospace and automotive - are traditionally the premier users of modelling and simulation software, a trend which continues, according to MathWorks, partly because it is very difficult for them to design aircraft and vehicles without using this type of application. "It is more important in aerospace because you do not want a plane [to get into] the sky until you are sure it is safe," Martin points out.

Many large companies in the aviation industry - including aircraft makers such as Boeing - have steadily increased the complexity of their design computational modelling algorithms by running large scale, distributed simulations based on multiple different parameters set to ensure the robustness of the final model.

These are often Monte Carlo simulations that use repeated random samplings of the same job to provide more accurate probalistic results in less time than with previous hardware configurations. Standard definitions of Monte Carlo simulations describe a broad class of computational algorithms that rely on repeated random sampling to obtain numerical results. One typically runs simulations repeatedly to obtain the distribution of an unknown probabilistic entity.

"We are seeing an uptake in aviation where those physics can be used to do heavy calculations and allow people to access a lot more power on an on-demand basis, and particularly the cloud," reports Autodesk's Jonah Normand. "Customers use that to compress their time cycles, to get their results in a much shorter space of time."

The burgeoning conception of types of unmanned aerial vehicles (UAVs - or drones) has come to rely heavily on modelling and simulation software. Design engineers at flight control and robotics systems manufacturer Quanser used Maplesoft's MapleSim to develop the QBall-X4, a quadroctor helicopter with four motors and speed controllers fitted within 10in propellers enclosed in a protective carbon fibre cage, as they sought to get a better idea of the gyroscopic effects of spinning parts on the UAV's performance.

Elsewhere, Brazilian UAV maker Xmobots used SCADE Suite modelling software from ANSYS to build a drone for agricultural and environmental monitoring (including cattle counting) in remote areas of the country, using the software to test and verify prototype models, qualify documentation, increase quality and reliability of the system and ensure compliance with DO-178/ARP 4754 aeronautical standards.

Usage is not restricted to aircraft or UAV design. Lufthansa Technik, the aircraft maintenance, repair, and overhaul division of the German civil airline, uses ANSYS's software to simulate wear and tear of aircraft components - particularly jet engines - to prolong service internals and create new ways to repair used parts.

The company also signed a deal to supply European aerospace and defence manufacturer EADS with software which simulates structural mechanics, fluid dynamics, and electromagnetics on aircraft, helicopters, space launchers and defence systems in August 2012.

These assignments are taking M&S applications into pioneering directions that are also making unprecedented demands on the skills of software development and engineering teams. Again, HPC platforms are essential to ensuring that the models and simulations can be run in a time-efficient manner.

Automotive drives progress

Many companies in the automotive industry use M&S software to design and stress test engines and other components, as well as optimise the aerodynamics of a vehicle's chassis, looking at upstream and downstream drive forces. The ANSYS Simulation product is used heavily by car manufacturers engaged in Formula 1 motor racing for computational fluid dynamics (CFD) and wind tunnel testing.

MathWorks' MATLAB, meanwhile, was used to build a simplified model able to predict the performance of the Bloodhound supersonic car, which is aiming to surpass the world land speed record by travelling at over 1,000mph in South Africa next year.

Ferrari says that simulation systems from ANSYS aided the motorsport team past its competition to secure the FIA World Endurance Championship drivers', teams' and manufacturers' titles in the GTE Pro class and the teams' title in the GTE Am class, during the 2013 season. "ANSYS software helped to reduce overall drag, maximise downforce, and to optimise 'complex brake cooling systems," says Gilles Eggenspieler, senior manager for the fluid products line at ANSYS. "Simulation enables engineers to evaluate multiple product design ideas across a range of operating conditions, leading to enhanced product performance and integrity."

French-owned US 3D modelling software tools specialist Dassault Syst'mes SolidWorks also provides the design software which brought a range of vehicles from concept to production line - including electric cars, boats, motorbikes, and scooters, and even bobsleds.

Last February at the Sochi Winter Olympics in Russia, SolidWorks software helped engineers shave off the crucial milliseconds separating victory and defeat off the US bobsled team's times. Geoff Bodine and Bob Cuneo, the designers of the four-man bobsled Night Train 2, hailed SolidWorks as instrumental in their efforts to improve on their Night Train design.

Aware of the strict rules surrounding the design of sleds, Bodine said that he knew the 2D design tool they used for the first generation Night Train would not be enough to create the world's fastest bobsled. With bobsled speeds often exceeding 90mph, races are won by hundredths of a second meaning that the slightest of tweaks to the design can make huge differences.

The original bobsled's aerodynamics were optimised for the fast downhill track of the Vancouver competition in 2010, but the track at the Sochi Games has three uphill sections that require precise handling to generate the most speed out of the track's curves.

SolidWorks' Premium simulation tools were used to design and stress test various components included in dual-tracked vehicle the Shredder, a personal off-road vehicle for the leisure market made by BPG Werks.

As with the aerospace industry, the design of individual automotive components is equally diverse. Japanese car maker Mazda claims that it had increased the fuel economy while lowering the exhaust emissions of its SKYACTIV-D clean diesel engines, which power its latest Mazda 6 saloon cars using MathWorks model-based calibration software for engine controller design and verification.

MapleSoft's MapleSim was used by Renault's Mechanical Engineering Department to simulate lubricated mechanical systems, which predict engine seizure. MAN Diesel & Turbo in Germany using the mathematical software package for its calculations in the design of propeller hydraulic control systems.

COMSOL, meanwhile, has been working with Ford Motor Company to develop mathematical models of batteries for automotive applications. Thermal management is a critical issue in the design and development of Li-on batteries for hybrid vehicles. The increasing use of simulation modelling has been a critical development in altering the way in which battery engineers have met challenges of thermal management, says COMSOL, by providing new insights into all the physical phenomenon occurring inside the battery.

Software in operation

Other vertical sector industries starting to increase their use of M&S software include industrial equipment, construction, medical devices and consumer products, with the software also finding a place in academic research projects and computational finance or even financial modelling.

Industrial automation equipment traditionally used very simple control systems made up of programmable controllers but are now using far more complex electronic and electrical systems.

MathWorks collaborated with agricultural vehicle manufacturer CNH (New Holland) to build an automatic trailer filler system, IntelliFill, which uses a 3D camera to guide the flow of crops into a trailer towed behind a tractor, allowing the driver to operate the harvester while filling a trailer accurately even when visibility is limited due to dust, poor light, or other airborne debris.

The CNH engineering team used the MathWorks Model-Based Design approach, MATLAB, and Simulink software to create a closed loop simulation of the trailer filling process. This helped build an on-board system that processes images taken by the camera and formulates instructions, which are then dispatched to the machine controller, with IntelliFill now fitted as standard to New Holland's FR9000 series forage harvesters. Using MathWorks meant the project could be completed in far less time compared to using C development language for the same purpose which would have meant CNH developing their own debugging tools.

Modelling and simulation software is also being used to test the strength and durability of materials used in the manufacture of various consumer electronics devices, including computers and mobile phones. "It is about making sure the product lifecycle is respectable, recyclability, and stuff," comments Autodesk's Normand, "so they stick the numbers into a major CAD system and make changes to the model to give the manufacturer an idea of the cost of the material and its environmental impact, almost instantly."

Construction scenarios involve not only architectural design but also analysis of how various systems can be fitted into large manufacturing facilities, for example, calculating energy requirements and optimising heat and cooling distribution across the facility. Autodesk has worked with UK data centre design specialist Sudlows, for example, which uses its Simulation 360 package to tools to verify the architecture it recommends for its customers, design new data centres, and retro fit olds ones.

Additional reporting by Edd Gent.

Further information

Share |

Hardware helps : HPC on-demand opens CAD/CAM to SMEs

Increased use of high-performance computing (HPC) clusters, which pool the resources from multiple high-powered computer systems, is an important hardware factor helping to extend the availability of M&S software into smaller organisations that would not normally be able to afford buying in sufficient processing power for high-definition 3D design and simulation requirements. Some of this change is as a result of CAD/CAM software vendors joining with HPC specialists. Autodesk is partnering Intel and French server and supercomputer vendor Bull in the UberCloud HPC Experiment, to explore potential HPC processing as a service for CAD/CAM applications for SMEs, for example. The UberCloud HPC Experiment began in July 2012, and as of December 2013 could boast more than 700 participating organisations and concerns (such as Russian computer-aided engineering firm Fidesys) engaged in experimental projects based on its facilities.

Over 124 teams have been formed, working on a wide variety of projects in digital manufacturing, life sciences, and Big Data analytics applications covering subjects such as computational chemistry, fluid dynamics, drug design, combustion, climate and weather change, and finite element analysis, and using crowdsourcing and social networking tools to aid collaboration.

ANSYS, meanwhile, offers additional HPC packs for use with its modelling software, which offers bundled access to demand parallel processing resources based on hundreds or thousands of CPU and graphical processing unit (GPU) cores. This is alongside HPC job management features and remote access to simulation jobs running in computer cluster and cloud environments.

M&S software solutions player SolidWorks does not directly offer access to its software through HPC time shares or public environments, but is examining the feasibility of doing so as customer demand increases. The company plans to support "virtualisation, remote usage and access anywhere" in future products.

Standards slip in: compliances entering the build

Data sets often include specific regulations or compliance rules which insist that equipment used meets certain specifications or safety requirements - the International Standards Organisation (ISO) 26262, which applies to the design and testing of safety-related systems on road vehicles, is just one example, and modelling and simulation software plays an important role in making sure that prototypes are compliant with these standards from the beginning.

MathWorks, for instance, offers model-based design consulting services for ISO 26262 projects, thus allowing automotive engineers for its MATLAB and Simulink products to build embedded systems that comply with the standard. It offers an IEC certification kit for both ISO 26262, and the EC 61508 functional safety of electrical/electronic/programmable Electronic Safety related systems (E/E/PE or E/E/PES), which contain certificates, workflow references, test suites, and tool qualification for artefact templates that can be applied to various components within its software.

MathWorks' Simulink Verification and Validation modules check that the size of the complexity of the model does not exceed given limits around block counts, maximum subsystem depth, and stateflow constructs; it can also be applied to the European IEC EN 50128 standard, which specifies critical software for railway control and protection systems.

ANSYS SCADE System also provides support for EN50129 safety-related applications for the railway signalling industry, alongside other engineering processes with high dependability requirements which include ISO 26262 and Aerospace Recommended Practice (ARP) 4754 guidelines for the development of civil aircraft systems.

Related forum discussions
forum comment To start a discussion topic about this article, please log in or register.    

Latest Issue

E&T cover image 1407

"Even the smallest of creatures in the most far-flung places around the world are getting wired up for tracking"

E&T jobs

E&T Marketplace

The essential source of engineering products and suppliers.

E&T podcast

Tune into our latest podcast

iTunes logo

Subscribe

Choose the way you would like to access the latest news and developments in your field.

Subscribe to E&T