The June 2015 issue of Harvard Business Review includes an article, Beyond Automation by Thomas H. Davenport and Julia Kirby, about the effects of automation on the workplace. The article is not at all about mesh generation, but rather on how your future employment prospects could be either jeopardized or enhanced by increasing automation.
Its interesting conclusion is that you should not look at automation as a simple replacement for current processes performed by humans, but instead use automated methods to collaborate with humans to augment their capabilities and result in new and better ways of working. Or put another way, simply replacing humans with computers is a near-term money saving measure, but using computers to relieve humans from repetitive tasks so they can apply their talents for more creative solutions results in greater long-term benefits.
If you use Pointwise or have looked at our website, you can see that philosophically we follow the augmentation approach, using automation to relieve engineers from many repetitive meshing tasks so they are freer to apply their judgement to get the best results for their problem. This is different than automatic meshing approaches that focus on completely relieving people from meshing tasks, even if the results are not that great, in order to get the short term benefit of replacing a human who was tasked with creating a mesh. There are other parallels between computational fluid dynamics (CFD) meshing automation and points made by the article’s authors that I’ll discuss more below.
The article focuses on career strategies and how you can ensure your job in a world where automation is replacing many workers. Five strategies for coping with increasing automation are identified: Step Up, Step Aside, Step In, Step Narrowly, and Step Forward.
The first two strategies, Step Up and Step Aside, are just to avoid the automation problem completely. With Step Up, you are in senior management as a person with the overall big picture and able to integrate all that information to make decisions better than any computer. Step Aside is the approach for highly creative people using intuition to satisfy customers in non-rational, emotional ways that no computer can match. Since these are essentially avoidance strategies and not ways to directly deal with automation, I’ll skip them for this discussion.
Step In requires knowing about the automation, what it uses as input, what it does with that input, and what it produces. In the CFD world, the input is the grid and boundary conditions, the physical models and numerics control what the automation does with it, and gobs of flowfield information are the results. If you want to Step In, you will know enough about this whole process that you can correctly set it up and then step in when needed to correct errors that naturally occur.
As an example the article provides a recent case,
Perhaps you saw a 2014 story in the New York Times about a man who had just changed jobs and applied to refinance his mortgage. Even though he’d had a steady government job for eight years and a steady teaching job for more than 20 years before that, he was turned down for the loan. The automated system that evaluated his application recognized that the projected payments were well within his income level, but it was smart enough to seize on a risk marker: His new career would involve a great deal more variation and uncertainty in earnings.
Or maybe that system wasn’t so smart. The man was Ben Bernanke, a former chairman of the U.S. Federal Reserve, who had just signed a book contract for more than a million dollars and was headed for a lucrative stint on the lecture circuit. This is a prime example of why, when computers make decisions, we will always need people who can step in and save us from their worst tendencies.
In CFD there are plenty of cases of where boundary conditions were set improperly, the grid was messed up, or a solution was not fully converged and the resulting flowfield, while looking beautiful in color pictures, was completely wrong. These types of problems are insidious, and you need knowledge and experience to examine your computational results and have confidence in them.
A more obvious automation problem shows up in automatic mesh generation. What if you replaced the human doing the meshing with a one button push approach? Push a single button, and the mesh is automatically built. This is great in concept, but what happens when it does not work? Maybe it does not provide enough boundary layer resolution or maybe it does not produce a mesh at all. How can you Step In? Push the button again?
That’s where the Pointwise approach to automation comes in. We automate repetitive meshing tasks, but we always provide a manual way of completing the task for those cases when the automation either fails to produce a result or gives you something less than you desire. Pointwise gives you the ability to Step In. As expressed in the original article, we look at automation as a way to augment the capabilities of the person to produce a better end result and not simply as a cheap replacement.
The next strategy is Step Narrowly, where you become so much of an expert in a particular field that a computer cannot match your knowledge. You find a niche and pursue it with focus and passion.
Another story from Beyond Automation,
In Boston, near the headquarters of Dunkin’ Donuts, a reporter recently peered into “the secret world of the Dunkin’ Donuts franchise kings.” One of them, Gary Joyal, makes a good living (if his Rolls-Royce is any indication) by connecting buyers and sellers of Dunkin’ Donuts franchises. As the Boston Globe put it, Joyal “uses his encyclopedic knowledge of franchisees—and often their family situations, income portfolios, and estate plans—to make himself an indispensable player for buyers and sellers alike.” So far he has helped to broker half a billion dollars’ worth of deals.
Could Joyal’s encyclopedic knowledge be encoded in software? Probably. But no one would make enough doing so to put a Rolls in the driveway. It’s just too small a category.
You might say that until recently everyone involved in computational fluid dynamics was Stepping Narrowly. Unless you were a subject matter expert the field was inaccessible. That has changed with growth of commercial CFD software and the great improvements in ease-of-use. Now there is plenty of room in CFD for people who want to Step In, but that does not completely eliminate subject matter experts. There is still a great need for people who understand the complex physics behind particular flow problems and can guide others in how to solve those problems using CFD.
Turbomachinery applications of CFD are a great example of this. Efficiency of rotating machinery has progressed so far in the last 50 years that designers are working to eke out tenths or hundredths of a percent in performance gains. If the CFD processes used to analyze these machines have variability of even a single percent, it could completely hide the performance effects designers are trying to achieve, making it useless. For that reason, most turbomachinery manufacturers have strictly controlled CFD processes governing mesh type, topology, and resolution, turbulence models, solver numerics, and convergence criteria to ensure the repeatability they require to make their results meaningful.
Step Narrowly also fits with the Pointwise philosophy. Our Glyph scripting language can be used to control all or a part of the meshing process. Engineers following the Step Narrowly path can use Glyph to design a meshing process that meets required mesh type, topology and resolution for their class of problems and does so in a repeatable fashion. Many of our customers use Glyph to develop complete end-to-end, templated scripts for meshing their products. Compressors, turbines, submarines, torque converters, and refrigerated display cases are just some of the products that have been meshed automatically using Glyph scripts.
That does not mean CFD experts following Step Narrowly have to develop Glyph scripts that provide complete start to finish meshing of a geometry. There are useful scripts that only perform one part of the meshing process or only check a mesh that someone else created to make sure it meets company standards. The script described in this Pointwise webcast makes a copy of a grid curve (a connector in Pointwise terminology) then stretches it to fit between two points. This is a common repetitive operation the script greatly simplifies.
There is definitely room to Step Narrowly with Pointwise.
That brings us to Step Forward, which is building or enhancing the next generation of automation. As Davenport and Kirby say, “It’s still true that behind every great machine is a person—in fact, many people.”
For example, at the E. & J. Gallo Winery, an executive named Nick Dokoozlian teams up with Hendrik Hamann, a member of IBM’s research staff, to find a way to harness the data required for “precision agriculture” at scale. In other words, they want to automate the painstaking craft of giving each grapevine exactly the care and feeding it needs to thrive. This isn’t amateur hour. Hamann is a physicist with a thorough knowledge of IBM’s prior application of networked sensors. Dokoozlian earned his doctorate in plant physiology at what Lohr informs us is the MIT of wine science—the University of California at Davis—and then taught there for 15 years.
How does Pointwise fit with Step Forward? One way is through collaborations like the one described in the anecdote above. We are subject matter experts in CFD mesh generation and Glyph scripting. Many of our customers are subject matter experts in their field and not in mesh generation, but they want to automate the meshing process for their company. We have collaborated with customers on projects to automate meshing for everything from rocket engine turbopumps to fully-appended submarines with great success. Our meshing experts work with company experts to create tools that non-experts in the company use to automatically create meshes that conform to company best practices and do so in a consistent, repeatable manner.
There is another way to Step Forward with Pointwise. We are continuously improving our software, adding new meshing techniques and streamlining and improving existing techniques. If you have expertise in and a passion for CFD meshing, maybe there is a place for you on the Pointwise team. We do not have any specific job openings right now, but we can always find a place for the right talented person with enthusiasm for meshing.
Why You Should Love Augmentation
Davenport and Kirby say that employers should love augmentation, but often don’t because of an automation mindset. Their article summarizes,
For augmentation to work, employers must be convinced that the combination of humans and computers is better than either working alone. That realization will dawn as it becomes increasingly clear that enterprise success depends much more on constant innovation than on cost efficiency. Employers have tended to see machines and people as substitute goods: If one is more expensive, it makes sense to swap in the other. But that makes sense only under static conditions, when we can safely assume that tomorrow’s tasks will be the same as today’s.
I think that paragraph describes the Pointwise philosophy toward automation pretty well. We want to use automation to relieve you from mundane, repetitive meshing tasks, so you can creatively solve problems and come up with new ideas and innovative solutions.
Early in my career I worked at a large aerospace firm where we joked that management’s approach to project staffing was just to decide how many “engineering units” to place there. They tended to look at engineers as identical pieces that could be substituted for each other and did not consider variation in skills or technical talents between people. A natural extension of that is to consider automation as a replacement for employees. That works for “engineering units” when you don’t care whether those engineers are innovating and coming up with new ideas, which won’t matter as long as your competitors are not innovating and coming up with new ideas.
But today’s world is more competitive than ever, and innovative approaches and new ideas are required just to keep up. Automated, push-button approaches – when they work – are great for cutting the expense of doing what you are currently doing, but they don’t do much to help you move successfully into the future. Using automation to augment your own capabilities leaves open many more possibilities for increasing capabilities and improving processes, whether it’s automatically approving mortgages or generating CFD meshes. For that reason, we will continue to make Pointwise more and more automated, but we will also make sure there is always a way for you to apply your creativity when the automation is not good enough.
thanks for such a nice synopsis