into the wind

You know it is really windy when you see pieces of tree flying overhead during the daily walk.  It wasn't easy keeping the normal pace going into the wind, but the trip back was easy.  Going back to sports it makes you think about the impact of wind on running and what the right strategy might be.  

 

A bit of physics will sort this out.  Consider a thought experiment where a runner starts off on a square course from leg 1 to 2 to 3 and finally to 4 that she runs at a constant speed s.  There is wind of velocity w that parallels two sides of the course and is perpendicular to the remaining two sides.  The amount of effort she has to spend to overcome wind resistance is proportional to the amount of power she must develop on each leg of the course.  I'll leave the physics to a footnote, but it turns out the power per lap is proportional to 4s( s² + w²).¹  Since w² is never negative, having any wind always forces her to deliver more power.  Her time in a single person race will always be slower if she is running at a maximum power level.

 

 

Now her strategy is clear.  When running with the wind at her back she should make sure she is running in clear air - that none of her trailing competitors can take advantage of her breaking the wind.  When she is running into the wind, she should be drafting faster runners to lower her apparently airspeed.

 

The reason we don't see dramatic difference in races with and without wind is that, at the speeds a human generates in a marathon, the power necessary to overcome air resistance is small - only a few percent that necessary to run.  It can make a difference comparing race times, but the disadvantage isn't enormous in most running events. 

 

Cycling is a different beast.  Since the power needed to overcome air resistance goes as the cube of your speed and cyclists are approximately not wind-cheating, racing benefits from any aerodynamic improvements.²  Improvements occurred over the years, but dramatic improvements came when experts in fluid dynamics began to study the problem.³

 

One area where you see an improvement large enough to make a difference in a race is in the design of the wheels.  Here is a high level paper from a state of the art wheel maker discussing why a racer would want to do this along with details of their design.  It isn't very technical, but it does give a sense of the amount of work that is done combining computer models and real world testing in low speed wind tunnels.  It shouldn't come as a surprise that such wheels can be very expensive. A few thousand dollars a pair is common and some will shrink the bank account by over six thousand dollars.

 

An important point is the computer tools are available.  Getting time in a wind tunnel is a different matter, but computer modeling has become important.  The important thing is you need engineers who are knowledgeable and can build accurate models.  Tools like this can give a false sense that you know what you are doing and it is possible to do more harm than good.  

 

Education and experience are critically important

 

Which brings us to the "Industrial Internet"

 

Recently several people asked me to comment on GE's announcement of their new initiative.  Here is the note I sent:

 

A few people have asked me about this since it was announced.  It turns out I have some deep experience as one of the designers of the mask lab information system project at Bell Labs in the late 1980s.  This is really hard to get right and hinges on success in several key areas.  It has and is being done by a few companies and requires a non-trivial effort (I've seen estimates that it consumes about 25-35% of the cost of building an IC fab line these days - of course they require it to make their process work as so much is at the leading edge of what the tools and process can do).

a bit of commentary...

GE has made a big deal about the "industrial internet".   I don't have deep knowledge of their specific program other than their PR documents (which are too high level and shallow) and high level articles like this one from  Technology Review 

http://www.technologyreview.com/news/509331/an-internet-for-manufacturing/

This sort of monitoring and analysis is not new.  Science has been doing it for a long time as fundamental part of understanding experiments.  Starting in the 60s this began to become computerized and it was common in particle physics from the mid 1970s.  I was involved with its adoption in integrated circuit manufacture in the mid to late 1980s.  Places like Intel, Bell Labs and IBM were trying to capture rich information from silicon processing.  It turned out to be an expensive task (the Bell Labs effort, which impacted mask making at two locations, involved a dozen people full time and probably $3-$4M a year for a five year period - the efforts at Intel and IBM were at least ten times that size, but one that was cost justified.  Many components were required to make it work: lots of sensors, a flexible database, good programmers, experimental scientists (physicists and chemists), mechanical and electrical engineers, and *really* good process engineers.  

Figuring out what was noise, understanding the sensors and finding good process engineers were the most difficult parts - along with finding people who were actually willing to change the process.  Even then there were many elements that couldn't be captured and leading edge places like Intel create carbon copies of fabrication lines to minimize variation in process.  

Hewlett Packard, Microsoft, Intel and others attempted to understand how to do this more generally to create a product that could be sold to manufacturers.  There were some large problems - many of which trace to understanding the process at a deep level. I'm aware of at least two approaches that involved anthropologists along with the physical scientists.

It was very clear you don't mix a network (of any kind), database and sensors and expect magic.

I hope GE does well, but there are a lot of issues genericizing this that range from computer science to mechanical and electrical engineering to process control and sociology and anthropology.  It is very easy to build a money sink if you don't get all of this right.

Experience has taught me there is a lot of experience that lives in the production lines that is largely untapped by the engineers and managers who created the processes in the first place. It needs to be underscored that sensors need to be understood deeply - I'm continually impressed with the lack of education in this area in most computer science courses.  The people who traditionally do these things are production engineers, but outside of a few areas the field is almost dead in the US.⁴  I find it exciting that GE is working in the area, but producing anything that is reasonably generic is will be a hard multidisciplinary problem.  Any organization that wants to tap the potential benefit needs to understand their process and measurement deeply - and that may not be a bad thing.

 

I'm at the end of my allotted time, but need to mention a wonderful development.  

The maker movement where people are building things that tend to lace computation with the real world is growing.  People are learning how to use sensors and microcontrollers and cheap computers such as the Raspberry Pi are accessible to kids in middle school.  

At the same time about a half dozen people have told me they intend to get their amateur radio license this year.  Understanding radio at a practical level is something I would recommend to anyone who wants to know a bit more about how the modern world works.  I'm going to hazard a guess that the average ham radio operator knows more about the fundamentals of radio than the average wireless executive.  

Perhaps we are at the beginning of something great. Maybe there is a natural hobby for many who can make contributions to the "Industrial Internet" and beyond as we move closer to the real world with sensors and computers.  And perhaps a few will even be instrumenting their sports and mixing that with a bit of critical thinking for a deeper undersanding.  It gives me a lot of hope - much more than using computer science and networking for selling things.

 

__________

¹ OK - a bit of physics.  First note that speed does not have a direction - velocity has a direction and magnitude, speed just a magnitude.  First consider an earlier post where I show the power needed to overcome air resistance is proportional to the the cube of the head wind.  The force of the wind goes as the square of its velocity, so power for the runner to overcome wind resistances goes as s * w² , where v and w are both vectors. In this case the wind direction has been chosen so we can ingore the vector nature, so s and w will just be the scalar components.

So running with the wind at her back, our runner needs to develop power proportional to s (s - w)²  and going into the wind her power needs to be s (s + w)². 

The crosswind case is a bit trickier.  The wind now comes in at an angle and the magnitude of the the resultant vector is  (s² + w²)^.5.  The magnitude of the force is the square so the power on each crosswind leg is s(s² + w²).  

Now we can add up the total power per lap.  It is proportional to s(2s² + 2w² + s² + w² + 2sw + s² + w² - 2sw) = 4s(s² + w²).  

 ² Cycling at commuting speeds has a cost, but it isn't significant until you face a headwind that exceeds 15 mph.  Commuters can comfortably pedal away in an upright position that would be impossible at 25 mph.  It is possible to dramatically improve the aerodynamics of a human powered vehicle by enclosing it in a shell.  These velomobiles can be pedaled by normal humans at 25 to 30 mph for long distances that would tax an Olympic cyclist on a conventional racing bicycle. 

³ The same can be said for swimming - remarkably the equations are similar.  Water is a fluid that is about 800 time denser than air.  

⁴ Production engineering is alive and well in Northern Europe, Japan and China.

 

__________

Reicpe Corner

 

The Super Bowl is in a few days.  Try to make some healthy snacks.  These are too simple.  You can roast almost any root vegetable like the first recipe and try other produce like sweet potatoes and even cauliflower (which is great when roasted with the same hot sauce you use for buffalo chicken wings -- something like Fred's hot sauce)

 

Roasted Carrots

 

° slice carrots into thin - say 1/4 to 3/8 inch thick slices and toss in olive oil and a bit of salt.  

° roast for 20 to 25 minutes at 400°F

 

 

Roasted Chickpeas

 

Ingredients

 

° 15 oz can of chickpeas

° 1 tbl olive oil

° 1 tbl curry powder

° 2 tsp cumin

° 1 tsp chilli powder

° 1 tsp coriander

° 1 tsp non-iodized sea salt

 

Technique

 

° Wash the chickpeas and dry off on a paper towel

° toss them with the oil and half of the salt

° roast at 400° F for 35 to 45 minutes until golden and borderline crunchy. Don't burn!

° let them cool to room temperature

° combine all of the spices in a large bowl and toss the chickpeas in with the spices