2016-04-27

Which company is number one on the CAE market?

It probably comes as no surprise that ANSYS tops the list with almost $1B in revenues and a market share of around 20 percent.

What may be unexpected is that MathWorks captures second place with a 15 percent share, ahead of well-known PLM giants such as Dassault Syst mes and Siemens PLM.

The latter pair are the subject of many column inches in the media. The same is true for ANSYS. In contrast, MathWorks leadership of CEO Jack Little and scientific lead Cleve Moler keeps a lower profile.

Instead of creating a market buzz around their products first MATLAB, and later Simulink they took a longer road to success by introducing MATLAB into the academic sphere in the late 70s and early 80s.

Cleve Moler used the first version of the solution when he taught numerical analysis to students at Stanford. This turned out to be a smart idea: the engineering students liked the solution, handled their study assignments in through it and then when they went out into the workforce, they wanted to use the same software. This philosophy is in line with the idea that the best way to predict the future is to invent it.


FROM $500 TO $800 MILLION. Jack Little and Cleve Moler (right) are the men behind MathWorks flagship products MATLAB and Simulink. Moler is the mathematician and computer programmer specializing in numerical analysis who invented MATLAB, which is a numerical computing package, to give his students at the University of New Mexico easy access to Fortran libraries without writing Fortran.

In 1984, he co-founded MathWorks with Jack Little (who reprogrammed MATLAB in C) to commercialize the program. Today, this software and Simulink have more than a million users and bring in $800 million. In February of 1985, MathWorks sold 10 copies of their mathematical computer software MATLAB to the Massachusetts Institute of Technology (MIT). This was the first ever business deal for MathWork s software, and it generated $500 in revenue for the newly-formed company. More than 30 years later, MathWorks products have more than a million users, bring in about $800 million in revenue and the company employs over 3,500 people.

It s a success story, and prospects look good. Why? In a world where model-based design in general, and embedded software, sensor and recognition systems, automatic code generation for programming industrial controllers in particular are gaining ground, MATLAB and Simulink[1] are crucial sub-PLM tools. Both are fundamental tools for research and development and used throughout the automotive, aerospace, communications, electronics and industrial automation industries. They are also used for modeling and simulation in fields that are becoming increasingly technical, such as financial services and computational biology.

MathWorks was working in simulation-based design before many other simulation companies and their customers recognized the importance and power of this approach for improving the effectiveness and outcomes of complex product development, explained CIMdata VP and analyst, Stan Przybylinski.

With the growing importance of smart connected products in a wide range of industries, simulating their behavior early in the product definition becomes essential to creating build-able products that meet increasingly complex requirements, Przybilinski continued.



USED BY LOCKHEED MARTIN SKUNK WORKS. Available with the company s Release 2014a, Simulink Real-Time enables engineers to build, test, and run real-time applications from Simulink models on dedicated target computer hardware connected to their physical systems.

Lockheed Martin Skunk Works recently used Simulink Real-Time during its end-to-end design, analysis, simulation, testing and deployment of the X-56A[2], a research aircraft built for the Air Force Research Lab s Multi-Utility Aeroelastic Demonstrator program. X-56A s communication interfaces, signal conditioning, mode logic, navigation and closed-loop flight controls were all designed in Simulink and prototyped and deployed using Simulink Real-Time. The main idea behind MATLAB which is basically a technical computing language is that it can accelerate the pace of discovery, innovation and product development. It offers a programming environment designed for algorithm development, data analysis, visualization and numeric computation.

When used together with Simulink, a graphical environment for simulation and model-based design of multidomain dynamic and embedded systems, MATLAB can secure and verify that end-system performance will live up to the product s requirements.

What we do is give engineers the tools to manage the complexity of projects by focusing on function and integration, rather than implementation (i.e. manual coding). By building models in MATLAB and Simulink, one can easily verify the requirements and functionality by simulation. The models can also serve as a common reference for the different project groups’ collaboration and can be used for things such as design reviews, claims Yngve Nygren, managing director at MathWorks Nordic & Balticum and EMEA Indirect Channel. Additionally, MathWorks produces nearly 100 products for specialized tasks such as data analysis and image processing. So far MathWorks portfolio has turned out to be a successful concept. Except for a flat revenue curve between 2014 and 2015, the company s growth has been solid and continuous.

During 2015, revenues landed at $800M, a figure which puts the company in second place when it comes to revenues on the CAE market. The average growth of the Simulation & Analysis (CAE) market which, again according to CIMdata, ended up with 3.9 percent due to the strong US dollar. If reported in constant currencies, the growth would have been approximately 10.7 percent, which is important to keep in mind when evaluating 2015 figures. Here are CIMdata s estimated Simulation & Analysis revenues for 2013-2015:

2013

2014

2015

$4.314B

$4.652B

$4.835B

7.1%

7.8%

3.9%

The Simulation & Analysis market generally shows one of the fastest growth rates among all the tools sub-sectors in PLM.

This reflects the growing importance of Simulation and Analysis as part of the product development lifecycle, and the need to acquire and use validated simulation solutions, asserted Przybylinski.

We think that the growth is partly due to expansion of existing S&A applications in validation on the right side of the systems engineering Vee and the move to more front end analysis as part of product development on the left side of the Vee, Przybylinski continued. In addition, market leaders have expanded their portfolios through acquisitions over the last several years, creating more opportunities for cross-selling in previously separate customer bases.

So, which are the toughest competitors? The answer depends on the market in question and the application.

National Instruments is one company that is often mentioned, but other companies such as Dassault Syst mes, with their emphasis on Modelica, can also be seen as competing across some applications. Other competitors include Mathematica, MathCAD (now owned by PTC) and Maplesoft. But as with most things in the PLM space (in which CIMdata includes S&A), Microsoft Excel is also used by many organizations.



AN INSPIRING MOMENT. When IBM introduced its first PC, model number 5150 (pictured) on August 12, 1981, MIT-trained control engineer Jack Little realized the possibilities of using MATLAB and the PC for technical computing. He reprogrammed MATLAB in C and added M-files, toolboxes and more powerful graphics. A couple years later, in 1984, he was the driving force when MathWorks was founded, based on that solution. He is still the CEO of the company today.

In principle, all physical relations can be expressed in partial differential equations. This means that the more relationships there are to be expressed, the more calculations are needed. In the 1950s and 60s, mechanical calculators and slide rules were used to perform these calculations. Since performing calculations at the manual level was extremely time consuming, the industrial significance was not particularly meaningful. Admittedly, there were already computers in the 1940s capable of numerical calculations, but these machines took a considerable amount of time to work through advanced problems.

For example, the solution to a system of linear equations with 15 unknowns could take 48 hours. Today, any PC can solve the same operation in a fraction of a second. The secret behind the development of these tremendous time savings are the algorithmic programming languages that began with Fortran in the late 50s. Unlike the hardware-machine of the 40s, it was now possible to formulate algorithms in a more mathematical way. In the 1970s and 80s, computerization and new numerical calculation models made it possible to take advantage of virtual models in product development activities. This became increasingly important, especially in the resourceful aerospace and automotive industries. This created a broader market with a rapidly growing demand, where the farsighted developers of these algorithms could find profitable provisions for their products.

One of the most forward-thinking players following this course was mathematics professor Cleve Moler. He came to Caltech as a freshman in 1957, initiating a successful academic career mainly focused on numerical analysis. He later became a professor at the University of New Mexico, teaching numerical analysis and matrix theory. During this time, he used Wilkinson and Reinsch s handbook for Fortran to produce EISPACK. This was followed by LINPACK, a package of Fortran programs for solving linear equations.

I wanted my students to be able to use these new packages without writing Fortran programs, so I studied a book by Niklaus Wirth to learn about parsing computer languages, Moler wrote in a 20-year anniversary article in 2004. In the late 1970s, following Wirth s methodology, Moler used Fortran and portions of LINPACK and EISPACK to develop the first version of MATLAB.

MATLAB proved to be very useful from an engineering point of view, covering topics such as control analysis, signal processing[3] and an emphasis on matrices. In 1984, Moler founded MathWorks together with Stanford and MIT-trained control engineers Jack Little and Steve Bangert. Little was the principal developer for one of the first commercial products based on Fortran MATLAB. When IBM announced their first PC in August, 1981, he quickly anticipated the possibilities of using MATLAB and the PC for technical computing.

He and Bangert reprogrammed MATLAB in C and added M-files, toolboxes and more powerful graphics.

MathWorks was born.

KNOWS WHAT SHE S TALKING ABOUT. We re working with some of the smartest customers in the world, says Loren Shure. She has been with MathWorks for almost 30 years, and makes the point that MATLAB and Simulink users tackle some of the most advanced product development problems there are.

We are working with some of the smartest customers in the world, said MathWorks Loren Shure, when I recently met with her in Stockholm. She presently holds the title Consulting Application Engineer[4]. This is a bold statement, but she knows what she s talking about. Shure has been with the company for almost 30 years and, as a matter of fact, she was the third employee in the newly-founded MathWorks. Over the course of these years, she has been occupied with developing the software, and when she talks about smart customers, she is making the point that MATLAB and Simulink users generally tackle some of the most advanced and complex problems associated with modern product development.

Obviously, Loren Shure and the product development team at MathWorks did a great job with the software. It s not a coincidence that MATLAB and Simulink have been a crucial part of some of the toughest technological challenges in product development history. For example, they were involved in the development of autonomous systems long before the automotive industry seriously began to think in terms of self-driving cars. The rovers Spirit and Opportunity of the Mars Exploration Rover mission (MER) back in 2004 are an excellent example. Landing the rovers safely at predefined sites, after a 320-million-mile-journey, was equivalent to throwing a dart in London and hitting the bull s-eye in Los Angeles. At this distance, it takes ten minutes for data sent from Mars to reach Earth.

One of the problems with this is that you can t perform any direct interventions in order to adjust deviations during the landing sequence. The vehicle had to be intelligent, in the sense that everything had to be controlled by an autonomous system. The landings, and subsequent independent rover missions, were the culmination of three-and-a-half years work and an expenditure of more than $800 million. Failure was not an option, so the engineers at NASA s Jet Propulsion Laboratory used MATLAB and Simulink for numerous phases of the mission including navigation, data analysis and the entry, descent and landing (EDL) system design. Jennifer Petrosky and Frances Flynn describe its involvement in the MER mission as a case study:

Prior to the rovers seven-month journey, navigation engineers had to prove that the rovers would make it to the Martian Gusev Crater and Meridiani Planum landing sites. Using data from previous missions, they built a simulation tool with MATLAB that took various attributes of the Martian surface into account to confidently predict that the chosen landing sites could be reached.

Prior to the Mars rovers (Spirit and Opportunity) seven-month journey, navigation engineers at NASA s Jet Propulsion Labratory had to prove that the rovers would make it to the Martian landing sites. Using data from previous missions, and the latest data from both the Mars Global Surveyor and Mars Odyssey, they built a simulation tool with MATLAB that took into account various attributes of the Martian surface to confidently predict that the rovers could reach the chosen landing sites.

One of EDL s systems engineers was responsible for ensuring that the landing craft would withstand disturbances in the Martian atmosphere from two to three kilometers above the surface to the landing site, and determining how the onboard systems would react to those disturbances. The terminal descent analyst, a long-time MATLAB user, developed a statistical program using MATLAB to propose, develop, test and implement two new onboard EDL systems.

The two new systems were manifested on the flight vehicles as TIRS and DIMES. TIRS, the Transverse Impulse Rocket System, was designed to add a last-second attitude correction to protect the landing craft’s airbags from self-induced excess horizontal velocity resulting from multibody excitation due to wind shears and gusts.

DIMES, the Descent Image Motion Estimation System, used pictures to detect prevailing wind and residual trajectory velocity effects on the system. The craft’s onboard computer processed images to correlate features properly and aid the Inertial Measurement Units (IMU) with a horizontal velocity measurement.

Using MATLAB, they were able to test, tune and implement the onboard descent systems, which told the rover which TIRS rockets to fire and when to fire them, based on images taken of the surface and IMU measurements acquired. So, MathWorks products are big in space, but what about on earth?

It is the same here, specifically in industries such as automotive. As the vehicles during the last decade have been equipped with increasingly sophisticated, highly electronic and software-controlled functions, the use of MathWorks and similar systems has almost exploded.

ON DISPLAY. The Scania Driver Support system analyzes the performance of the driver and suggests how to further improve their driving. The system was developed using MathWorks tools for model-based design. Truck manufacturer SCANIA is one example of this. In the transport industry, the progress towards sustainability and autonomy is first and foremost business-driven. Step-by-step increased electrification, advanced driver assistance systems and autonomous transport solutions reduce the CO2 footprint and help maintain profitability in an industry with decreasing margins.

Advanced driver assistance systems also fill the gap between fully human-controlled vehicles and completely autonomous vehicles. Inefficient driving techniques can increase carbon emissions and raise fuel consumption by as much as 10 percent. For a company with 20 trucks, each being driven 120,000 kilometers (75,000 miles) per year, this translates to some 200 tons of additional carbon dioxide emissions and 66,000 Euros in added costs. Scania s driver support system provides immediate feedback to drivers via a dashboard console. The system helps improve driving techniques, leading to greater fuel economy, safer driving and less wear and tear on brakes and other parts. Now in production on Scania R-Series trucks, Scania Driver Support was developed using MathWorks tools for model-based design.

Modeling the system in Simulink enabled us to define the architecture, visualize the design and run simulations to debug the design at an early stage, explained Jonny Andersson, lead engineer at Scania. With Embedded Coder, we generated code for early real-time prototypes as well as for the production system. As a result, we refined the design in the model instead of in low-level code.

This example is only the tip of the iceberg; there are many, many more where MATLAB, Simulink and the like are used in the vehicle industry. For instance, this includes highly advanced systems that are usually summarized under the term Active Safety such as ABS (Anti-lock Braking Systems), ECS (Electronic Stability Program), and TPI s (Tire Pressure Indicators).

ACTIVE SAFETY. TPI (Tire Pressure Indicator) was introduced to the US in the Audi A6 during 2008. The solution was developed by Swedish NIRA Dynamics using MATLAB among other software. TPI is a software solution which determines the tire inflation pressure from the wheel speed signals and does not require any in-wheel pressure sensors or RF components. It is clear that MathWorks software plays increasingly important roles as product digitization continues to grow at an exponential pace. As discussed above, the general trend is that model-based design and simulation are growing and the use of these tools has spread explosively from being a tool for cutting-edge research organizations to being used even by medium and small enterprises.

One of the drivers of this is big business tendency to outsource not only individual component development and production, but also to buy ready-made systems to facilitate final assembly. For example, automotive manufacturers today don t buy only the bare dashboard of a car. Rather, they prefer to have a complete package, with software, electronics, gauges, lights, etc. in other words, a system of functionalities. And they want to have it in one piece, ready to mount and connect.

This phenomenon can be observed not only in automotive companies such as Toyota, General Motors, Volvo or BMW. It has also spilled over into the consumer-market producers such as Ericsson, Whirlpool and others. The point is that these system and product development efforts are now largely being subcontracted. The result is a need to build up digital-based product development chains, which contain not only the usual suspects in terms of integrated PLM/PDM tools, but also the important components of advanced numeric calculation and simulation.

This means that product developers and manufacturing engineers need tools that can ensure that the functionalities of software connected to electronics and sensors runs as intended. Another aspect is that those who buy these systems want product verifications as, for example, simulation results.

Signal processing algorithms are becoming increasingly complex, as is the contextual programming. Many signal processing algorithms consist of structurally parallel data flows and iterative computation for parallel and sequential data points.

Because graphical processing units (GPU s) consist of thousands of smaller, more efficient cores designed for parallel performance, GPU computing is well-suited to accelerating the computation and simulation of signal processing algorithms. However, programming for GPUs can be complex.

The combination of MATLAB and GPUs provides a technical computing language as well as the computing power of GPUs, and enables engineers to perform computations on powerful GPUs using the MATLAB environment. By changing only a few lines of MATLAB code, engineers can accelerate an algorithm with a GPU, speeding a radar system simulation by hundreds of times. The picture shows an Angle Doppler map of the simulated clutter return generated with Phased Array System Toolbox.

This is a brave new world, and MathWorks is growing with it.

Yes, indeed, said MathWorks Loren Shure. And our customers do amazing things with the software. They do sort of two major different kind of activities. One is using a model-based design flow, where they try to design and describe dynamical systems so that they might be able to do extra things with them. Maybe control something in them.

So for example, in the case of the Mars rover, they used Simulink models to figure out how the rover was going to land. They couldn t test that out so they did many, many simulations, trying things out under different conditions and making sure that they had things set up so the landing would be secure, Shure continued.

Then we have another kind of customer and these tend to be more on the MATLAB side who are very interested in, again, possibly modeling a system, but in many cases just trying to understand some information. They have a lot of data and maybe want to come up with a theory of what is happening and MATLAB lets you go through a scenario like that: try a bunch of options and say, Well, could it be because of this or because of that, and run simulations compared to data. MATLAB and Simulink are like a pair of twins how do they work together?

So, MATLAB is a textual language; you write code in it. It s meant for doing linear algebra, which is the language used for science and engineering. So it s a very natural way to express the constructs that you think of for mathematically understanding the world, said Shure.

Now, what Simulink does so brilliantly is that it lets you think about the world in a different way. It thinks about the world as a dynamic system and it lets you pull together pieces that fit together in the way you think about them. In your car, for instance, you have the axle and it s connected to the wheels, and so you can literally do that in Simulink. You can describe your system in it, model the dynamics by watching it run under certain conditions, Shure added.

Having these two modalities means that many times there are good reasons to combine them. Since they are part of the same eco-system, there s every reason to include some of the MATLAB analysis in the middle of a model-based design workflow to gain better insight or simply a better model.

In December 2014, Siemens Automation joined MathWorks Connections Program. This network contains more than 400 complementary third-party solutions and includes products and services that complement MATLAB and Simulink. That said, it is important to bear in mind that tools such as MATLAB and Simulink are part of a larger context PLM which means there are several aspects to consider from a product lifecycle perspective.

One consideration is the fact that usage of these products tends to expand into the manufacturing area, outside of product development. MathWorks has created the MathWorks Connections Program to build an ecosystem of product development partnerships, one piece of which manages product lifecycle effectively, to help create seamless development chains with smoothly flowing product data. This is the goal of Industry 4.0, for example. MathWorks Connections Program is available to third-party organizations that develop and distribute complementary products and services based on MATLAB[5] and Simulink[6].

Siemens Automation is one example of a company that joined this network. In December 2014, it announced an expanded relationship with Siemens. The purpose was to make it easier for customers to take advantage of model-based design[7] for developing and implementing complex control algorithms. As a result, control engineers can now generate code from Simulink and deploy it to Siemens SIMATIC S7 modular PLC controllers and SIMATIC WinAC RTX software controllers. Subsequently, advanced control strategies can be designed and tested using simulation to prove out concepts before further validation is performed on PLC hardware.

An increasing number of control engineers in the machinery and automation industries use simulation and automatic code generation to supplement programming industrial controllers, helping them to move faster from ideas to implementation and to expand their lead in innovation, said Philipp Wallner, industry manager for industrial automation and machinery at MathWorks.

We are working together with a lot of vendors to work out how our tools can fit more naturally into the PLM lifecycle, says MathWorks Chris Hayhurst.

Putting Intelligence into Systems

This offers an expanded scope, but what about PLM in general, and how MathWorks products interact with the different environments offered on the market including systems such as Siemens Teamcenter, Dassault s ENOVIA or PTC s Windchill? I spoke to MathWorks Chris Hayhurst, consultant manager, on this topic.

Well, the PLM system is vital for keeping in play the whole complexity of the development process, he replied.

MathWorks tools have historically been used as point design tools. So when it s necessary to do some design work on a component or a small module within a system the artefacts necessary for that design can be brought out of the PLM system. MathWorks tools can then be used to do the whole design cycle, the exploration of the data and the algorithms bringing that together and producing production software, said Hayhurst.

That is then wrapped up and put back into the PLM system, and the lifecycle is maintained through the PLM system. Now that interaction is maturing and continously improving, we are working together with a lot of vendors to work out how our tools can fit more naturally into the PLM lifecycle. So you can now see that MathWorks tools are more involved in the systems engineering level, as well as at the component level design, so you can do studies bringing together whole sets of components and work out how a virtual integration could prove that a system is actually on track and ready for deployment. In a perfect world, this PLM integration would have been 100 percent ready, up and running. But as with everything to do with progress, it s still a step-by-step process.

It is clear that MathWorks has embarked upon that journey and will eventually reach a high level of seamless integration. The company s software is spot-on when it comes to the requirements of cutting edge technology, and as a result the demand for their solutions is growing fast.

What do you regard as the key growth factors in terms of this year, 2016? I asked Chris Hayhurst.

We are driven by our customers, he replied. We work with a wide range of industries. At any one time, some of those industries are really exploding and growing. There s huge movements going on at the moment in the automotive industry. The rise of ADAS (Advanced Driver Assistance Systems) are promoting a huge upsurge in the computing content of cars.

MathWorks is all about enabling engineers to really put intelligence into their systems. So a change like that in one particular market can move the whole business for us, our relationships with our customers and at any one time there are two or three of our industries that are just at that point where there s something really interesting happening with big investments, Hayhurst said.

MathWorks plays in a market segment where its products are in line with what is needed. There is no doubt; the future looks bright, a window is open, and MathWorks is ready to climb inside.

References

^ MATLAB and Simulink (www.engineering.com)

^ X-56A (www.lockheedmartin.com)

^ signal processing (se.mathworks.com)

^ Learn more about this title (www.linkedin.com)

^ MATLAB (www.mathworks.com)

^ Simulink (www.mathworks.com)

^ take advantage of model-based design (www.engineering.com)

Show more