How did you master control engeneering? your own project? just books?

how did you master control engeneering? your own project? just books?

Not him but bump

I didn't

Nobody masters it without realize that they really can't.

All of those help.
Reading books, thinking about shit. Making simulations, writing a simulator. Making your own shit (make that drone, robot, whatever).

Who hard is to find control jobs in the process industry? Is it worth going for a master degree in control?

How*

this, you just get good enough with sisotool. ;p

I wouldn't say i've mastered it, but building my own projects really helps. If you look at it from an engineering perspective the only thing that really matters is knowing that a problem is solvable and how to implement its solution. Understanding it completely and theoretically is something mathfags worry about. Get a bunch of sensors and motors and whatever process you're interested in controlling, start with something simple like a PID motor controller, getting this to work well requires nonlinear filtering of the input and output signals, polynomial interpolation or something similar, and all sorts of tricks like using PWM to control motor power, voltage dividers to get a tachometer signal into the voltage range for the ADC, basic knowledge of programming microcontrollers and checking signals with an oscilloscope... also don't waste any time learning analog methods like implementing transfer functions with op-amps, nobody ever uses this stuff anymore

Im a mechanical engineer, but i'm messing around with this stuff on my own. I think the future of control engineering will merge with parallel computing. In the near future these control loops will be implemented on digital micro controllers with millions of cores, this will replace all analog methods entirely.

also laplace transforms are a meme, learn state space representation and linear algebra.

I'm interested in this too, thinking about a phd in control theory. but not sure i'm smart enough, and want to get out and make money in engineering.

Control systems videos on youtube. There's this guy who does an excellent job explaining the material. I believe his name is Brian Douglas.

These videos helped me a lot.
How is Scilab compared to Matlab or any other tool?

This.


But generally it's not the most complex subject.
Infact once you get the hang of it it's quite enjoyable.

>Laplace transforms are a meme.

MechE dude here too, but with a math background.

U wut m8? Any state space representation can be represented via a transfer function and any transfer function can be represented by a SS representation. In fact, it's easier to get the controller form from the transfer function since all I have to do is draw the signal flow diagrams. I can also get observer form easily, too.

Now, if you're talking nonlinear. ...then yeah, laplace transforms and the TF are memes. Also impossible.

I've heard that Laplace Transforms are barely used in real control applications. Is it true?
I personally think it's really frustranting, since as an undegrad I learned control based on Laplace Transforms and transfer functions.

It really depends. This may sound ridiculous, but it depends on your style.

I come from a mathematical background (double degree -- Math & MechE), so I've always found transfer functions to be intuitive. For me, I glean a lot of information from a transfer function that can help me out when I design the simulated system.

For example, a transfer function naturally tells me where the poles are in the frequency domain. It also tells me if it's BIBO stable or not, where it becomes unstable, etc. It's very, very useful for me. It also helps when I need to find the specific gains K for the system. Finally, if I have a very good estimate for the transfer function, I can use Simulink, pick only three blocks to model the system. Those three are:
>Transfer Fcn block
>Scope
>Input block

Now, there are other engineers who don't need to use it or have so many years of experience, they can more or less understand the the system without it. The state space representation works best for them. That isn't inherently bad and I find that people who have a natural mechanical engineering mindset are better at it. My leanings have been more mathematical (even pure) and my interest in control theory is actually borne out of my immense interest (read: actual research interest and why I'm funded as a PhD student) in dynamical systems, where control theory is merely a subset of dynamical systems.

Now, I'll be honest -- the vast majority of my control theory work has been primarily academic in nature. Nonetheless, when my PI gives me a small assignment on a contract that is associated with the DoD or government, I use the same concepts and techniques I learned in the classroom plus extra. I lack the hands-on guidance, which other MechEs seem to have down to a tee.

>I personally think it's really frustranting, since as an undegrad I learned control based on Laplace Transforms and transfer functions.

Generally, most applications of control theory irl are a hodge podge of academic techniques, intuition and experience/trial & error. You should know that irl, nonlinearity is the norm and 99% of the time, the stuff you model/simulate is approximated from nonlinear phenomena to linear phenomena. They do undergrads a disservice by teaching linear only, never touching nonlinearity. Nonlinearity is where you learn the essence of control theory because it tells you when to approximate, how to approximate, where to check for stability/instability, etc.

All linear models are essentially derived from nonlinear models. They're "dumbed down" versions of it, actually. Your goal when you simulate/model an inherently nonlinear system is to find what are called the equilibrium points, model it linearly around them and factor in other issues (sensors, PLCs, motors, etc. List goes on).

I hope I've answered your question(s) in detail.

>Understanding it completely and theoretically is something mathfags worry about.
I'm an engineer too, but a sound math base which you can build upon and knowing how something works "in the background" aka knowing the theory can really help with some stuff and makes you a better problem solver. After all everything is based on a mathematical model.
This.

I apologize for asking, this may seem a little off-topic, but I am a ChemE student that's quite interested in control theory. Taking a Process Control Dynamics course this year and would really like to pursue this field more closely.

Would I get better results if I just switch to ElecE instead? Or should I just pursue an MSc in control theory after my ChemE undergrad?

No need to apologize. To clarify, I am
I really can't offer any advice for your situation. Control Theory is mainly an EE/MechE field and honestly, it's flavored generally toward EEs a bit more because a lot of MechEs shy away from the intensive mathematical aspect of CT. For undergrad, I took a lot of EE coursework to go alongside my MechE coursework because they offered tons of courses with a clear CT slant. For example, in a given semester, they would offer
>Matrix Theory course (if you don't know Linear Algebra, you don't know shit about CT. This course did nothing but talk about applications of linear algebra and transformations. Very, very applied).
>Linear Systems & Control Theory (EE course)
>Nonlinear Systems with CT Applications (EE)
>Digital Signal Processing (very good EE course with close ties to CT)
The list goes on and on. I exhausted the CT courses for MechE. MechEs who are interested in control theory are usually those with math backgrounds (like me) or those involved in interdisciplinary studies (robotics, mechatronics ... also like me).

That being said, I've encountered some problems about chemical plant processes (in fact, very standard examples talk about it), so there's obviously a need for it.

The best advice I can give you from my perspective is that you should take some elective EE classes dealing with CT and, if you feel the "pull" of it, then go for an EE major. But honestly, it would be very unique for prospective companies to look at your resume/transcript and say to themselves
>wow this person is ChemE but has a lot of knowledge in controls engineering -- both as a ChemE and EE person!
Before you make that plunge of changing majors (and it is a huge plunge), dip your toes in a little. Test the waters.

Thanks for answering. I was afraid that transfer function approach I learned would turn out to be useless, but now I'm kind of relieved.

I understand that most of the real life models are nonlinear. However, in my control course I was taught to linearize nonlinear models through Taylor series. Is this a valid approach to deal with nonlinearity, or irl that's not how it's done?

All ChemE programs have at least one course on process control, there are several ChemE books about the subject. Naturally, books and professors tell that ChemEs are best suited for control jobs in the process industry because they have more knowlodge about the processes. Now I don't know if this is really true, or if professors say that because they are biased towards ChemE, and EE can actually do the same job (or maybe better).

It is a valid approach, but irl, you're usually not given neat and clean nonlinear models like from the textbook. Also, they taught you that doing the taylor series around the equilibrium points is generally nothing more than finding the partial derivative at the eq. point, correct?

I should also clarify that although I personally use TFs, not all MechEs use it. State space representation works just as well and many mechEs can glean information from SS representations alone if they're able to derive them.

Thanks for the help, I've become quite disillusioned with chemical engineering over the past year. It seems like we are just glorified factory workers and the job seems hideously monotonous.

At least in ElecE there is some form of active development going on, maybe even a more design-oriented role upon completing graduate work in ElecE. Not to mention the job market, ChemE is sucking ass right now.

Read a book on functional analysis, learn about its applications to optimisation, learn enough EE and programming to create a circuit that will self-optimise.

Pic related, but I didn't really master it untill had to develop assignments as a TA and realized how ugly non-linear systems really are and the importance of the more theoretical side of control optimization.

EE or ChemE? ChemE is hard right now because of the oil price, EE is still relatively healthy.

Never do an engineering PhD unless you want to gun for academia. You're much better off getting a masters, if you only want money it's nearly always better to not do grad-school at all.

>Now, if you're talking nonlinear. ...then yeah, laplace transforms and the TF are memes. Also impossible.
>I've heard that Laplace Transforms are barely used in real control applications. Is it true?

When you think about the feedback loop in the OP you can imagine an arbitrary non-linear process G_p, while G_p doesn't have a laplace transform, you can still have some background simulation (or real data) in that block which take an input signal and produce an output single. Otherwise the math works as it does normally although you are more limited in the analysis.

Even systems that are actually truly linear have a noise in the output etc. and you use the usual methods to handle this. Barring more advanced processes, the practice is to fit a linear transfer function as a first approximation then design the control system and then use on-line tuning to get the final control parameters.

>They do undergrads a disservice by teaching linear only, never touching nonlinearity. Nonlinearity is where you learn the essence of control theory because it tells you when to approximate, how to approximate, where to check for stability/instability, etc.
Half of our first semester dynamics was literally just "linearize this non-linear system", it really scares me that there are departments out there that only teach linear systems to undergrads.

EE control engineering is more focussed on intrumentation and signals while ChemE control engineering is more focussed on the processes.

Pick your poison, the math stays the same.

>EE or ChemE? ChemE is hard right now because of the oil price, EE is still relatively healthy.
ChemE. I'm going to finish the course in less than two years, so I'm not really considering change to EE. How about control jobs in non oil related process industries? Do they exist in a reasonable amount?