YouTube: https://youtube.com/watch?v=BT_JUbpZ1lQ
Previous: Intro to Big Data: Crash Course Statistics #38
Next: Big Data Problems: Crash Course Statistics #39

Categories

Statistics

View count:330
Likes:32
Dislikes:2
Comments:8
Duration:11:02
Uploaded:2018-11-15
Last sync:2018-11-15 16:30
Engineering, like life, could really use a lot more cheese. This week we are looking at a cheese factory in Toronto and what it can teach us about process control systems. We’ll explore feedforward and feedback systems, and see how integrating them both with the final check of cascade control creates a system made to handle uncertainty the world throws its way.

Crash Course Engineering is produced in association with PBS Digital Studios: https://www.youtube.com/playlist?list=PL1mtdjDVOoOqJzeaJAV15Tq0tZ1vKj7ZV

Global Weirding with Katharine Hayhoe: https://www.youtube.com/channel/UCi6RkdaEqgRVKi3AzidF4ow

***

RESOURCES:
https://www.improbable.com/2018/01/10/the-100th-birthday-of-murphy-the-murphy-of-murphys-law/
http://www.abc.net.au/science/k2/trek/4wd/crasht.htm
http://www.proflow.ca/case-studies/process-control-misapplication.html
https://www.explainthatstuff.com/howtoiletswork.html
https://www.marineinsight.com/tech/heat-exchangers-on-ship-explained/
https://www.controleng.com/single-article/applying-heat-exchanger-control-strategies/81538b26c934f76ca70fde5d7b2181f6.html

***

Crash Course is on Patreon! You can support us directly by signing up at http://www.patreon.com/crashcourse

Thanks to the following Patrons for their generous monthly contributions that help keep Crash Course free for everyone forever:

Sam Buck, Mark Brouwer, James Hughes, Kenneth F Penttinen, Trevin Beattie, Satya Ridhima Parvathaneni, Erika & Alexa Saur, Glenn Elliott, Justin Zingsheim, Jessica Wode, Eric Prestemon, Kathrin Benoit, Tom Trval, Jason Saslow, Nathan Taylor, Brian Thomas Gossett, Khaled El Shalakany, Indika Siriwardena, SR Foxley, Sam Ferguson, Yasenia Cruz, Eric Koslow, Caleb Weeks, Tim Curwick, D.A. Noe, Shawn Arnold, Malcolm Callis, Advait Shinde, William McGraw, Andrei Krishkevich, Rachel Bright, Mayumi Maeda, Kathy & Tim Philip, Jirat, Ian Dundore
--

Want to find Crash Course elsewhere on the internet?
Facebook - http://www.facebook.com/YouTubeCrashCourse
Twitter - http://www.twitter.com/TheCrashCourse
Tumblr - http://thecrashcourse.tumblr.com
Support Crash Course on Patreon: http://patreon.com/crashcourse

CC Kids: http://www.youtube.com/crashcoursekids
Engineers often worry about laws.

Not necessarily the legal system kind, although most engineers have to consider those as well! What I mean is laws, like the laws of physics.

Those laws describe limits on the sorts of things that are possible; for example, there’s no way of getting around the conservation of momentum that we know of! But outside of physics, there’s one law that keeps many an engineer up at night: Murphy’s Law. Murphy’s so-called ‘law’ is more of a tongue-in-cheek proverb, named after aerospace engineer Edward A. Murphy. Simply put, it says: “Anything that can go wrong will go wrong.” That makes it sound like all attempts at engineering are doomed. But we’re going to find a way to use Murphy’s law to make us better engineers. [Theme Music] Murphy’s “law” isn’t a literal statement.

Your calculator isn’t going to spontaneously combust every time you use it. But keeping in mind all the different ways mishaps can jeopardize an engineering process, as Murphy’s Law tells us they can, is a good reminder to stay vigilant. Even if many parts of a process are in flux or changing, you need a system to keep things stable and reliable.

That same system should also prevent mistakes from spiraling into a major disaster! To do this, engineers use what’s called process control: an automated system that takes a look at what’s going on in your process, and makes adjustments based on those observations to keep everything on track. The aspects of the system your process control looks at are called parameters, while the outputs that get changed as a result of those observations are called controlled variables.

To see all this in action and understand why process control matters, consider cheese. A few years ago, a cheese company based in Toronto ran into trouble when some of the vats they used for storing milk started having problems. The vats needed a certain amount of milk in them for the cheese-making process.

Like many industrial processes, precision is what allows high quality products to be manufactured with consistency. And no one wants bad cheese! Too little milk in the tank, and it would affect production further down the line, perhaps leading to a loss of the entire production run.

Too much milk in the vat, and some of it would start to spill over. At first, that might not seem like a big deal. Even if it did spill over, some of it would be directed into the drains leading to the sewer, removing it from the factory floor.

The rest could be cleaned up and, as in the case of too little milk, you might need to write off this production run. But that’s not the worst of it! Cheese is made using bacteria.

If the milk containing that bacteria overflows from the vat and into the drains, it could end up in the sewer systems. And at the end of the sewer line would be a waste treatment plant that uses its own bacteria to treat wastewater! If the bacteria in the cheese began to kill off the plant’s bacteria, it would jeopardize the whole system, perhaps even threatening the water supply of an entire urban area, like the city of Toronto.

This sounds like exactly the sort of catastrophe Murphy’s Law tries to warn us about! Thankfully, process control is exactly the sort of thing that prevents both major and minor problems like these. What was happening with the cheese factory was that the inputs to the system of milk vats needed to be controlled to maintain the level of the contents inside.

In the language of process control, one of the parameters involved in making the product, in this case, the level of milk in the tank, is sensitive to the controlled variables of the process, which are the amounts of all the ingredients pumped into the tank. The failure was that the system that automatically adjusted the flow rate of the ingredients in response to the measurements made in the tank wasn’t working as intended. As we’ve mentioned, process control uses measurements of a process’s parameters to make changes to its controlled variables.

That might sound a little abstract, but various types of process control are currently in use around us, making the world run smoothly. When you set the thermostat in a building, you’re essentially programming a process control system to keep it at a particular temperature – say, 25°C. That’s the set point – the number that represents the target output or operating state you want a process to achieve.

For example, after you set it, the thermostat then turns the furnace or the air conditioner on or off, to heat up or cool down the environment and maintain it at 25 degrees. The controlled variable here would be whether the furnace was on or off, which has a direct effect on the parameter you’re interested in, the temperature. Other parameters you might want process control to be sensitive to are things like the pressure in an oxygen tank, or the force being applied to something.

You can even use it to control a mixture’s chemical properties, like acidity. Because of its broad uses, process control is everywhere. It’s a major consideration in chemical, electrical, industrial, and mechanical engineering, just to name a few.

The beauty of it is that along with avoiding big mistakes, process control lets you create products to particular standards with consistently high quality and precision. Plus, monitoring and controlling things so carefully can help you find new ways to use materials and energy more efficiently. In modern control systems, measurements are taken as electronic readings from sensors, which are delivered to a computerized control system, called well, the controller.

The controller also sends signals to the machinery responsible for changing the controlled variables. This normally means the operation of things like valves or switches. There are two main kinds of process control to consider: feedback, and feed-forward – which, yes, is a real word.

To see this in action, let’s revisit our old friend the heat exchanger! Heat exchangers transfer thermal energy from one fluid to another to raise or cool the temperature of one of the fluids. In a shell-and-tube exchanger, you run one fluid through a series of pipes to exchange heat through the pipes, with the surrounding fluid.

Let’s say you’re using steam in the pipes to raise the temperature of oil to 200°C before it flows into an engine. At the input end for the steam, there’s an inlet valve that controls how much steam is entering the exchanger. The goal is to make sure the oil leaves the heat exchanger at a given temperature – 200 °C – that’s the set point.

The controller then adjusts the machinery to maintain the set point condition for the process – in this case, by controlling the steam inlet valve to maintain the oil’s temperature at 200 degrees. One of the ways you can do this is with feedback control. First, you put a sensor to measure temperature at the output end of the exchanger for the fluid.

In feedback control, the sensor will continuously feed temperature data back to the controller. The difference between the observed temperature of the output and the set point is called the error. The controller tries to minimize the error by controlling the inlet valve at the start of the process to increase or decrease the amount of steam entering the exchanger, depending on whether the error is negative or positive.

If the error is negative, that means the temperature is too low, so the controller will open the valve a little to let more steam through. If the error is positive, it will do the opposite. On the other hand, you could also use a feed-forward system.

In that case, you’d be measuring the input variables of the process. In the heat exchanger, the sensors would measure the amount of steam and fluid flowing in, and the fluid’s starting temperature. Then you’d model what you think the output temperature will be, based on those inputs.

In this case, you’d need to know things like the specific heat of the fluids and the heat conductance between the steam, the pipes, and the oil. The modeling is often the hardest part, but once it’s done, all the input parameters provide a decent estimate of what the output temperature of the fluid will be. What’s more, from the model, you can then work out the best flow rate for the steam, given all the other inputs.

In this case, the difference between the measured flow rate and the flow rate needed to obtain the set point temperature would be an example of what’s called the disturbance. The disturbance is the difference between what the input parameters should be and what they really are – like the output error in a feedback system. All the sensor data is fed forward to the controller to compare the modeled output temperature to the set point.

Depending on the disturbance, it operates the valve similarly as before, until the inputs are set according to the model to make the oil’s temperature hit the set point. On their own, both of these approaches have some flaws. In a feedback control system, you have to wait until something has already gone wrong in order to fix it!

If a lot of errors are creeping into the process very quickly, a feedback system might not respond fast enough to change the inputs. On the other hand, a feed-forward system relies on having to model everything going on inside the heat exchanger. You can’t be sure that your model is perfectly predicting the output temperature, and even if it’s pretty close, it usually takes a lot of work to get the model right in the first place.

In real life, often the most sensible thing to do is just combine the two approaches. The feed-forward controller can help you get the inputs as close to what you need as possible, while the feedback controller can correct for the flaws in the model by measuring the actual output temperature. You could even implement something called cascade control as an additional measure.

It helps you make sure that turning the valve influences the steam flow rate in the way you expect. To do this, you put in a separate controller and sensor that measures the steam flow as the valve is opened or closed by a certain amount. Otherwise, if the steam is at a higher pressure than expected, opening the valve might increase the flow rate of the steam by more than you predicted!

The cascade control system puts checks in place to prevent that from happening. With the feedback, feed-forward, and cascade systems working together, in what’s called an integrated approach, you can keep the oil’s temperature steady at the set point, despite all the little ways Murphy’s law could have made things go off the rails. You can apply the same thinking to the milk tanks in the cheese factory.

To keep everything perfectly balanced, the sensor in the tank was meant to measure the milk level, while other sensors measured the flow rate of ingredients. The controller would then adjust the flow rate as needed. But because the sensor – in this case a gauge in the tank – was misreading the level of milk, the entire system was being thrown off.

After fixing it to take accurate readings, the factory’s process control started working properly again. And the water in Toronto stayed nice and clean. So while Murphy’s law might paint a pessimistic picture of engineering, process control steps in as the linchpin that keeps everything from spiraling out of control.

Crisis averted! In this episode, we looked at process control systems, where automated controllers change process variables in response to measured parameters. We looked at how feed-forward and feedback systems minimize errors and disturbances, and saw how integrating them both with the final check of cascade control creates a system made to handle uncertainty the world throws its way.

Next time, we’re looking at how systems react to forces, when we delve into the world of statics and dynamics and how they affect all the structures an engineer might create. Crash Course Engineering is produced in association with PBS Digital Studios. Wanna keep learning?

Check out Global Weirding, which explores the intersection among climate, politics, and religion, hosted by climate scientist Katharine Hayhoe. Crash Course is a Complexly production and this episode was filmed in the Doctor Cheryl C. Kinney Studio with the help of these wonderful people.

And our amazing graphics team is Thought Cafe.