When Paleo men built tools for the first time in human history, they learned an important engineering lesson: build-test-repeat loop. The shorter this feedback loop, the greater the pace of innovation. And we live in a time when the pace is at its all-time peak, and we can surely say the pace tomorrow won’t be as slow as today. As a citizen of Bangladesh, this writer cannot ponder and ask what our place on this highway of innovation is. Historically, we have always been at the consuming end of innovation. Yet, recent developments in various technical fronts can give us an entry point to turn the table and start serving technological innovations to the world.
Current state of simulation technology
For most of the industrial age, the build-test-repeat loop was exceptionally long. The building phase was always the costliest and longest one, with numerous occasions when organizations brought a product to market too late or of inferior quality and paid the price due to the laggy nature of this loop. Because of these experiences, there was always a drive to devise ways to shorten this feedback cycle.
Since the dawn of computing, there have been efforts to incorporate computers in the design process so that part of this build-test-repeat loop can be done virtually, which can save both time and money. Aerospace and nuclear reactor design industries were the early adopters, with automobiles following suit quickly. Today, virtually all industries, whether running shoes or high-end cellphones, paper towels, or HVAC systems, invest deeply in simulation technology. Modeling and simulation have become the bread and butter for gaining key insights during design and testing.
The last 50 years have seen enormous growth in computational science, and ideas from academia have poured over to commercial products. Most successful commercial software lineages can be traced back to university laboratories. Although academia is still living on the bleeding edge and working on new challenges of simulation technologies, much of its output has matured enough to become stable tools for industrial application.
We have seen a continuous investment from companies to make these tools user-friendly, reliable, and efficient. In the last decade, we have experienced the appearance of digital twins, which can represent a physical product or process completely in a simulation environment. Digital twins have become a key enabler of Industry 4.0.
The open-source innovation
It might feel daunting and cost-prohibitive for small and medium-sized organizations (SMEs) considering adopting this new stream of technology, particularly those not from developed nations. Fortunately, most of these software have open-source alternatives with the caveat of limited customer support and not-up-to-the-mark UI/UX.
Many successful commercial products are built on top of some open-source products. NASTRAN and Code Aster are two highly capable open-source finite element simulation software, and quite a few commercial software are based on them. OpenFOAM can feature-wise compete with Ansys Fluent and Star CCM+. If we are looking for some first-principle-based modeling, FENiCS can be a really powerful tool. Also, not to mention, the Python ecosystem has a lot of tools that can offer the same capability as MATLAB.
So, for both academic institutions and industrial organizations, if they want to dip their toes in modeling and simulation, open-source software can be an alternative to long-term investment in commercial counterparts. Also, this is an opportunity for local academic institutions to collaborate with international academic institutions and make fundamental contributions to these open-source software.
The Cloud Model
Since the 1980s, there have been predominantly two kinds of simulation infrastructure: one using a node-locked licensed version of simulation tools on an expensive high-end desktop and the other using a relatively light desktop to build the model and then using the HPC cluster to run the models. Both cases require upfront capital investment for hardware and software, often running the tab as much as 500,000 USD for simple use cases. This is often an entry barrier for SMEs adopting modeling and simulation in their design process.
With the emergence of cloud-based computing, we are now experiencing two new business models: Software as a Service (Saas) and Infrastructure as a Service (IaaS). These models offer organizations the ability to overcome the entry barrier. Now they can start small and scale up as needed. This also frees up a valuable investment they can now divert to invest in technical development.
The commercial simulation software vendors are also pushing for these changes as this helps them to capture more untapped users. We are also seeing a shift to a subscription-based pricing model. The subscription model helps the vendors predict their future revenue consistently and helps the users to amortize the software cost over a period. We also see a push to offer low-fidelity and fast simulation products over the cloud, particularly targeting designers. Along with the market leaders like Ansys, Siemens, and Dassault Systemes, many start-ups like SimScale, Simerics, and OnShape are popping up and offering pay-per-use models for users that save them from upfront investments.
In Bangladesh, the thought of investing in modeling and simulation is likely to be shot down in an executive room meeting due to the lopsided cost-benefit curve. Even for large corporations, mid to long-term investment in such technical ability development comes with many risks due to factors like business climate, talent turnover, etc.
However, with these new pay-per-use and subscription-based models for hardware and software, Bangladeshi companies need not feel left out when it comes to leveraging modeling and simulation to innovate. They no longer need to funnel heavy capital investment for infrastructure and talent development. Starting small and organic growth of simulation footprint can be the new mantra for local engineering firms.
Innovation opportunities at home
You will often hear simulation tool veterans saying, “Garbage in, garbage out.” Indeed, computer models merely generate outputs based on the inputs given, and they rarely judge the quality of the input. So, if the models are weakly developed, the simulation results will be unreliable. There are many case studies where companies invested in the wrong modeling techniques and found the results unusable. That scarred experience left them second-guessing every later venture they did in modeling and simulations. So, investment in simulation tools will never bear fruit without investing in developing the users.
This opens a new avenue for collaboration between academia and industry in Bangladesh. Often, when advocates propose adopting modeling and simulation in the design and manufacturing process, executives get scared thinking that now they must completely change how things have been done so far. This is not how successful organizations incorporate simulation technologies in their workflows. It often starts with addressing low-hanging fruits. An organization that aspires to adopt simulations in its processes should identify small but rewarding technical design or manufacturing problems it wants to resolve using simulation. The next step would be to develop collaboration with academia, where the expertise is already available to develop models and interpret results.
In the past, the unavailability of hardware and software would have stalled the process here. But today, numerous open-source and low-cost pay-as-you-go services are available for most simulation problems, particularly in the low-fidelity domain. This is where academia should train current students with the first-principal modeling and the latest off-the-shelf software. Most software companies are willing to work with academic institutions to offer low-cost or free academic versions of their products, as this helps increase their user base.
Since commercially available simulation tools have reached a certain maturity, for any academic, it feels daunting at first to make any fundamental contribution in this domain. Bangladeshi academic institutions should take a more pragmatic approach to developing talent in this space. First, we need to overhaul our engineering curriculum by eliminating outdated coursework and introducing simulation courses. There should be courses that teach basic first-principle-based simulation techniques and utilize industry-standard open-source or commercial software. For example, mechanical engineering students should learn to develop models based on the heat equation or the laws of elasticity and take courses that teach modeling and simulation using Ansys or Nastran
The key focus should not be on reinventing the wheel but rather on understanding the principle of how the wheel works and then using existing wheels to solve some real-world problem. Academics should not confine themselves to their offices and wait for industry professionals to bring problems to them. Rather they should take an active role in reaching out to the industry and seeking out problems to which they can offer solutions.
Developing talent in any domain is a medium-to-long-term project. It would be foolish to expect any short-term gain from the investment. There should be a fast feedback cycle of solving small problems and using the results in the industry, which will develop the analyst community’s competence and confidence. As this progress, problems with incremental complexity can be tackled. On complex problems, there will be temptations to get it done by shipping it overseas. Undoubtedly, this will be a costly and harmful move for the long-term local talent development process.
Bangladesh has raw talent that can be nurtured and morphed into a capable high-tech workforce. Due to various shifts and congruence in technology and business models, a new opportunity has opened up, and we should not waste it.