Tuning A Control Loop - The Knowledge Board

[@failranch9542]

I could tell within the first minute that this guy is a natural teacher.

[@marcoaureliovillarrealdelv2765]

SIR, YOU ARE A GENIUS! A very practical way on physically implementing control loops. I have to say that I used to struggle a lot with such concepts in college. This video is gold.

[@daltonjameson1506]

You prolly dont care but does someone know of a method to get back into an instagram account? I was dumb lost my account password. I would appreciate any help you can give me

[@darthenx2585]

Your a genius man. Great explanation. You truly know exactly what your talking about. I would love to work on automation projects with you.

[@mstjerning8919]

Great video! Very concise and no nonsense :-) It might be worth mentioning that when deriving a model from a step-response (I call this a data-driven model), the step used for the response should be in the linear region of the plant/process. Take for instance a motor that can run between 0-1000rpm. To derive a linear model for this motor, the step should not be close to the lower (0 rpm) and upper (1000rpm) operating regions of the motor, due to non-linearities from e.g. inertial mass (shaft, gearbox, load). I would recommend performing an initial step of say 200rpm, followed by a step from 200 to 800rpm. This last step is used for the linear model. The low and upper regions for a motor tend to be non-linear and would not produce an accurate linear model. For validation I tend to check my stability margins by evaluating the gain and phase margins of the system (yes, I use a classic frequency analysis design approach). Typically I calculate these margins and measure them using an analyzer. In most cases all checks out, but you could risk running into instability due to resonances that are not apparent in the 1st order model. Cheers from Denmark :-)

[@erickmartinez6805]

Kevin, You are brilliant. You teach control as it is in real world. thanks.

[@DavidRead52]

Your first point is so RIGHT the number if times I have gone to breakdowns "Due to the controller" that is something physical not functioning correctly in the big bad world. The worst was the wrong sensor been installed from construction and after a lot of engineers looking at the controler over 3 years, then I go to site, doing your first thing and look at all the sensors and find they all were getting reading from a pressure tranducer that was the wrong type. Remember there is no such thing as a desk engineer. Very Good Video

[@user-yy6zv3ou3f]

How did you recognize it was the wrong type or what was wrong about it.

[@Gary-ts6dh]

9:15 - Wow! I've been to engineering school and I never made that connection! Approximating τ as Δ/4 will be extremely useful to me into the foreseeable future! Thanks.

[@emmyherregodts9082]

Although some people are very good at it: trial and error tuning should be avoided as much as possible. This is a very nice example to extract information out of process data for tuning loops with simple dynamics ..... However if you follow this instructions to the letter your proportional gain (Kc) will be most of the times not correct and you will not get the intended closed loop behavior. The proportional gain(Kc) should be based on the scaled process gain. So if you use in the calculations Kp*(Output_sf/PV_sf) then your result will be fine. If you use tool for this you have to give in the scale factors and the software takes care of everything. If you plan to do the calculations manually (which is not bad at all) then never forget to include the scaling.....Wrong (or no) scalings = Wrong tuning

[@jjhaak1240]

You understands it very well, it is indeed important and cant be neglected

[@Peter73sky]

I'm now closer to understanding PiD settings ,before I had always guessed.Thank you.

[@GiveForaSmile]

Simply an amazing video. It was well summarized and yet extremely informative, specially being just the tip of the iceberg. I am leaning on doing my senior project on PID systems based on what I just learned from you. Thank you

[@dougtowers]

One of the best explanations I have watched about a close control loop tuning. A master. Thank you for sharing with all of us.

[@petersmith3564]

Thanks for taking the time to share your knowledge. You have a knack for explaining a complicated process and breaking it down in a fashion that even a noob can grasp. Cheers!

[@socialcivilian2703]

Confident and Concise session. This knowledge is greatly appreciated. Good thing you had 5 boards at your disposal.

[@sampathjoshi1054]

Very practical and honest explanation from his dedicated experience! Received an realistic insight into the topic! Salute to you! ....Love from India!

[@josephpetersen8811]

This is the best PID video I’ve seen yet.

[@destroyo5457]

I knew I was watching a genius at work when he converted a delta symbol to a 4 on the fly at 9:15 😉

[@ajaydorasid]

Wow....That's the best explanation I've ever got about PID loops... Thanks a lot....

[@stanpikulski4007]

Kevin, you are a great teacher and engineer, THANKS!!

[@EndrChe]

This guy is a world class educator.

[@pardeepchhikara2170]

Thank you for taking out time for this wonderful explanation.

[@jaykesapalaran5434]

Very good teaching... Please also make a more video tutorial specs of PLC of what is about is in the marking .

[@iamhe999]

so where and what value did you set the the P, the I, The D?

[@nayrbsworld3048]

Crystal clear explanation.

[@hayfordankamah-cofie6687]

Good facilitator,very informative.

[@memethematics1873]

Fabulous Sir! Thanks a lot ABB service!🙌 Hope to get more informative videos at here!

[@zucontrol]

Sp validation is good. Also good to validate by load changing to see how fast pv return to sp.

[@HemantSingh-gz3kx]

Really great video on Tuning Fundamentals,,

[@farzadhagh8582]

wooow , no word can describe my thankness

[@francescobroers9018]

Thanks for the very informative and practical explanation.

[@stuartp2006]

Video games: Stormworks uses a Parallel PID generally, From The Depths uses a Classical. I think

[@FullCircleTravis]

I'm going to need more whiteboards.

[@masifamu]

thanks for this lucid explanation

[@awadelrahman]

That was really clear, practical and helpful. Thanks a lot.

[@kumudayanayanajith6427]

123445687359324876234 videos, 6476354812364129876491386 research articles summarized in to just under 22 minutes! Great video, Great explanation. Thanks ABB

[@grego.1418]

so much so quickly... excelent

[@dansrailways1]

An absolutely brilliant video, very well explained!

[@mohammedmousa3506]

ABB is the Best

[@davidrose9848]

Thank you for the info, it has definitely helped me understand the basic concept of tuning. Awesome video, thanks for sharing to the Automation world. By the way, i do love your VFD products. I have changed about 95% of our drives over to yours.

[@mohammadadil7763]

Excellent and most practical.

[@borishruskar5108]

Very, very nice. Bravo and big thank you.

[@andypeterson2682]

Excellent, very good explanation, my respects teacher ............... BRAVO ... !!!

[@goldilauks]

Thank you. That was incredibly helpful.

[@matheusviniciusribeiro7540]

This is a very good class

[@laurentporre4479]

Great ! But you didn't mention about the delay (dead time) potential issues, with a long delay, this tuning rule is much less efficient, but ok it works great in many cases.

[@davidcerce]

Laurent Porre Dead time is included in the process variable response. If it’s too long (compared to actual response) it means the valve is sticking (called stiction) in his first order example. Repairs would have to be made first and then perform new bump tests.

[@josemontenegro5679]

Excellent video, thanks !

[@gillv7659]

Similar to the way I do it, but on some occasions unable to perform even a manual step so watching the response adjust accordingly plotting values to see how the gain changes with various output settings. What about showing when to use a feed-forward and looking at the plant to determine the best way to control not as individual PIDs plus combing more than one process variable into a single control valve.

[@bramblebop1904]

That's a very good video. The main reason, it starts from zero, lol. I'm tired of books that hit you with nuclear-physics looking stuff on page one! Talk to me like I'm stupid cause I am! 😂

[@ingridepoch9143]

Super helpful thank you!

[@rudielamprecht6839]

Great video. I just need to find a process to tune now.

[@frogeman]

thanks! though it is the tip of the iceberg, it is much helpful to my knowledge at work.

[@krishnakurma4711]

very good explanation, Thanks

[@jhonjvare09]

I think the sensor should be after the actuator. I know this is only flow, but flow is affected by the valve. It is only a negative feedback system.

[@LeonardoSilva-oq4ey]

Thank you for share this with us. Excellent explanation. 👏👏👏👏😃

[@randyblythe6563]

Please create video for gain scheduling and also Feed forward

[@BrandonsGym]

This is really a nice video

[@AdamTrojnar]

that amazing :) so much important things explained in such easy way.

[@atiks]

Great video. Thanks for sharing.

[@michaelmatheson2279]

Thanks Kevin. Great stuff.

[@MarcoRotella]

Thanks for sharing this lesson. best regards

[@HellTriX]

Thanks. I felt like I was back in college learning stuff!

[@KianoushMir]

Very nice and informative, great work!

[@MedellinVol]

You're (or you are) a genius. I suppose you'd have to be a control nerd, as I am, to realize this, but excellent presentation.

[@outdoorelement6082]

Great video ! Extremely Informative ! Thanks for sharing.

[@phuocanh19811]

@14:22 We will do change the setpoint in closed loop. But what the initial value of the PID parameters

[@Micuamacu1]

Kevin, You remaked that is extremenly important to know the PID type (Standard, Parallel or Classical). However, in video I cannot see where this information is relevant. The fact you said to use standard, where it is used for calculation?

[@zanedzikonski4234]

Great explanation, I’ll have to watch again. Only thing I’m not understanding is changing of set point as it relates to time. For example using temperature as units if changing from 72 to 78 vs 84. The function of time changes but I’m not sure how. Any ideas? Will watch again...I love how many whiteboards you have btw 😆

[@LvtLoshi]

My man. Thank you!

[@rodericksibelius8472]

Wow it's just like designing a DC 'Resistor-Capacitor' Circuit.

[@krrishnaibm]

Nicely Explained!

[@elpirogoldfish69]

great video, good info, might need to sobriety check the camera guy tho..

[@marioachi2152]

I love this Videos! UFB Job!

[@chinox4567]

Good video and well explained!!

[@binhdaoduc6220]

thank you very much, it works

[@techzila5379]

Sir how we can design a PID controller for controlling the position of shaft which is supported by active magnetic bearing ?

[@mligua-zen527]

Nice sir, we need more real time problen

[@jbreiter56]

I get it now! Thank you.

[@kavyajt3814]

thanks a lot sir

[@phuocanh19811]

@14:47 could you please explain why the ratio is 1. What if the ratio is different than 1.

[@patrickjones7434]

Brilliant man

[@sathishkumar1023]

Amazing

[@RezaSimRacing]

amazing

[@m.abrarzahid8619]

8.42------- (i may be wrong but process value is our output of the system) How can you divide (del P.V with del output, for gain) Shouldn't it be ... ( del P.V / del input, which is our step input) Kindly correct me if im wrong.. anyone..

[@Micuamacu1]

Hi Kevin. I work for ABB New Zealand and I am interested in the training course single Loop tuning. Kindly, coud you please send me the training code or the link for apply the training?. thanks

[@deanswan6109]

well done...Thankyou

[@shuhratxomidov8315]

How do you do! Please! Could you present video's ABB lst300. How to calibrate it!

[@CheoChamarripa1]

Exeptional explication, could you teach us how to tune a second order closed loopr response?, I love you man!!!!!

[@feolender2938]

why are you talking about integral TIME? the integral part needs a gain? as does the derivative

[@davidcerce]

Feo Lender Derivatives is not used in first order curve loops.

[@denisseramirezvivas6352]

Excellent video! :)

[@ghulamhabib7810]

can u give lecture on differential pressure to steam flow cascade in distillation column . or suggest me any good book

[@emmanuelsheshi1553]

thanks so much sir...

[@bramblebop1904]

Where did you "go back to get a master's degree"? And what background did you have prior yo that, if you don't mind my asking.

[@jauharkhotami4073]

Really helpful

[@PudoRandonneur]

Great video. What I absolutely don't understand is at 15min16s: You say Td=0. But doesn't that value switch off the D-part completely and reduce the PID-controller to a PI-Controller?

[@andrewtillison9659]

Did you ever get an answer to this? This is bugging me as well. At this point, it's just a PI controller.

[@wereprobablyoutside4435]

I think you guys are correct. I have heard that most industrial systems do not use the derivative part, I am not sure why they wouldn't

[@sorcererdm200]

I dont know if you still wonder why they dont use derivative term, but here is 2 reasons why anyway :) ; 1- Derivative of error introduces a very high output initially if your set value is applied as step. For ex., you start the cruise controller in the car, controller sees a set point change from zero to 100km/h in just a few seconds; think how high would be the derivative of that and it will be fed now as an input to the motors with also Kp and Kı terms! 2- Noise. This one seems like a "ehh, they just say noise everytime they cant solve something" ( :)) but with a derivative term, noise can become a huge problem. Think of the magnitude response of the derivative term; it is Kd * w (|Kd*s| = |Kd*j*w| ), where Kd is the gain and w is frequency. As you noticed probably if you apply 5 Hz, derivative term would be 5Kd and if you apply 10Hz 10Kd. What if we apply 100kHz( a noise freq.) for ex.? then the noise would be multiplied with an enourmus constant while your control output is multiplied with something small. this reason alone can beat the advantages of derivative term in most cases. It is possible to get around both of these. You can activate the derivative bit later or change the term it derivates( directly output for example) for initial problem; for second one you may be able to add a lowpass filter before controller etc. but if you can meet the requirements with less component, time and effort why would you implement derivative and also deal with new problems.

[@davidcerce]

PudoRandonneur That’s correct, derivative is not used for a first order curve response loop.

[@juanericksolis]

Man..... you are my hero ;---;

[@Myvoetisseer]

I would also be that smart if I had that many whiteboards.

[@MuhammadSaleem-wv6pj]

Interesting

[@faboxbkn]

amazing! thank you.

[@SakakiDash]

Very informative video, thanks a lot. :D

[@perceptronsaber4479]

Very Good ; Thanks

[@djl6692]

Great video Kevin! It did leave me with a few questions. At 5:00 and again at 7:50 you describe bump testing and write out a change in the process variable based on the change in the output. My question pertains to what is "output". I keep thinking that we should be watching for a change in output based on a change of input... So I'm interpreting you to mean that the change in the output, is the output controller (presumably in milliamps from 4 mA to 20 mA), which serves as input to the control valve or the pump speed that would create a change in flowrate (to pick an arbitrary, hypothetical process variable). Presumably this output (my input) is a number that you can get from the control system historian (DeltaV/Foxboro/Siemens PLC/DCS) or taking a measurement with a multimeter across the two wires to the control element. Is that correct?The process variable is undoubtedly measured and also available from the HMI or historian. My next question pertains to the equations for Standard, Parallel, and Classical PID algorithms. You showed to solve for Kp and Tp and how some of the other parameters (Kc, Ti, Td) could be solved for if one assumes a 1st order model. However you present these 3 equations for standard/parallel/classical with variable S, and never define what S is. Are those expressions supposed to be equal to something? Are you supposed to set these expressions equal to zero to find S?