I swear, these videos are vastly better than most University professors. Understanding the benefits and drawbacks, as well as intuitively appreciating the underlying logic of engineering methods constitutes the critical learning foundation that is largely glossed over in academia. Keep up the great work producing this sort of content. It's greatly appreciated.
I am a control system professor .. You summarized what I had learned in years in few minutes .. Wonderful .. This presentation of such topic is amazing .. Please keeps it up .. I am enjoying your lecture
i m a mechanical engineer and i'm completely switching my subjects to control field courses (like 'control of vehicle dynamics', 'vibration control'...) just because of you. you are a great teacher and a great engineer . Ty so much, Brian, ! keep uploading videos man
@@utilizator1701 we actually know control systems, but for sure not much as a control/automatic engineer. anyway, if you are interested in automotive engineering or energy managment you find a lot of courses and elective courses focusing in simulink and control. That's based in what you like most and what kind of engineer you want to be.
brian ..... the reach of what u do is immense , programming, manufacturing, reverse engineering, engineering, biology, art (houdini and co) , research methodology etc . systems and their exploration and design should be thought as metascience , just like discrete math, information theory and probably cognition
Great stuff as always. Was really excited when I saw you were doing this series. One of your previous videos mentioned the place to spend the most time was in getting the model right so I’m looking forward to being able to improve mine.
Thanks! I wish system identification was the perfect fast solution to modeling. Unfortunately, it can still be time consuming and requires a lot of trial and error to get it just right. Hopefully, with this series you'll have a few more tricks though that you can use to improve your model! Good luck!
If I don't really know what the order of the system might be, I would first try to measure the frequency response function of the system, for example using tfestimate in Matlab (not to be confused with tfest) which uses Welch's method to average out the influence of noise. From this FRF data it will be a lot easier (after a bit of practice) to guess what a order of a model would approximate the system well, at least from the data used to obtain the FRF. I assume one of the next videos in this series might also cover this and also mention when the measured FRF can be "trusted" using the coherence (though that doesn't give a completely objective measure of the trustworthiness).
Great stuff Brian....still remember your spring and paint can analogy from back in the day.....one question though - this is something that I've seen controls engineer have misalignment on in citing some literature; when it comes to using System ID on closed loop system i.e. let say a PID based black box, this is common when it comes to using off-the-shelf components (motor drives etc), the method doesn't work well or is invalid in some people's view.....what's your take on using SysID in identifying a model of a system when a cascaded closed-loop is part of the system....
Hello @MATLAB, Hi Brian, I'm trying to fit a model to my data which is a time series sports data. I have 2 inputs vectors of under 10 representative features each, corresponding to the 2 teams state vectors at time t-1, my output is either a win or a draw ( so 3 possible output ) so I would like to output the probabilities for each possible output. I feel dumb but I can't wrap my head around that .. Let's say A vs B was the game at t-1, and A won the game, next week the game is C vs A, my goal here is to take the state vector of A ( as a said earlier it would be a vector or few parameters that represents A ( Number of goals, shot on target, ranking at the time of the game ) PLUS the fact they won, and feed that to the next step ( which is the input of C vs Agame ) and so on .. up to 5 steps. Maybe you could give me some tips to get started ? Thanks and a big thank you for your Control series they are amazing
Great video but I have a question. In the example, how can you tell by examining the system response that you don't need to include any zeros in G(s)? I have learnt that the poles tell if the system is stable or unstable, smooth or oscillatory etc but I don't understand what role the zeros play. Thank you and keep up the good work!
First of all, I like this series about controltools in matlab . I have a question .....actually it may sound silly. Why Mathworks has suddenly( not that suddenly...of course :D) started to make video tutorials about its tools although some of these tools were introduced more than 7 or 9 years ago? I am not aware of the whole process of product or software release , I know to sure tutorials take time to prepare....but not to the point that the company produces a tutorial for some of the tools inside its software after 9 years and maybe even more.
14:52 The rigid cable is connected to mass m and its displacement x1 is the response signal; pulley 1 is rigidly fixed to a wall whereas pulley 2 is linked to the wall with spring k. The center of pulley 2 undergoes displacement x2=x1/2. The second additive term in the differential equation should be ... +k*x1/2 instead of ... +k*x1/4, is it not? Thanks in advance for the response.
hi.. for a step response test for sample a heater.. what is the input and output of the system? is it the pid controller out (as input) and temperature (as the output)?
Great stuff as usual and perfect timing! Could you please include information: 1. How to setup experiments to obtain good quality data and some common input (excitation signals) to use and caveats 2. MIMO state-space estimation in this series?
Hi Brian that was a great video but I have a question, where can I locate the block you use at minute 9:46? The one called real system with uknown dynamics, I can't find it and it would be useful for my college project. Regards
That's just a regular Simulink subsystem that I labeled as "Real system with unknown dynamics". I created the dynamics in Simulink and then hid them away in a subsystem so the viewer couldn't see it. You can think of that block as whatever your real model or physical system is producing.
How can you know a mass and poulies system work from 1st principles .. and I didn't. What are the drawbacks of never using 1st principles to solve problems
I swear, these videos are vastly better than most University professors. Understanding the benefits and drawbacks, as well as intuitively appreciating the underlying logic of engineering methods constitutes the critical learning foundation that is largely glossed over in academia. Keep up the great work producing this sort of content. It's greatly appreciated.
Congrats to MathWorks for hiring Brian Douglas! That's the right move!
I am a control system professor .. You summarized what I had learned in years in few minutes .. Wonderful .. This presentation of such topic is amazing .. Please keeps it up .. I am enjoying your lecture
i m a mechanical engineer and i'm completely switching my subjects to control field courses (like 'control of vehicle dynamics', 'vibration control'...) just because of you. you are a great teacher and a great engineer . Ty so much, Brian, ! keep uploading videos man
I have thought that a mechanical engineer knows control system engineering.
I am confused in which fields is taught control engineering.
@@utilizator1701 we actually know control systems, but for sure not much as a control/automatic engineer. anyway, if you are interested in automotive engineering or energy managment you find a lot of courses and elective courses focusing in simulink and control. That's based in what you like most and what kind of engineer you want to be.
I have a playlist created for Brian Douglas (his own videos and the ones from MATLAB). All are outstanding and practical approach for control systems.
Awesome! Thanks :)
Do you how tool to make something showing brian to make grafics from formula
brian ..... the reach of what u do is immense , programming, manufacturing, reverse engineering, engineering, biology, art (houdini and co) , research methodology etc .
systems and their exploration and design should be thought as metascience , just like discrete math, information theory and probably cognition
When I see Brian’s video, I watch it. Amazing one, thank you!
Just a pity I didn’t have this series last semester when I was taking this course :)
Great stuff as always. Was really excited when I saw you were doing this series. One of your previous videos mentioned the place to spend the most time was in getting the model right so I’m looking forward to being able to improve mine.
Thanks! I wish system identification was the perfect fast solution to modeling. Unfortunately, it can still be time consuming and requires a lot of trial and error to get it just right. Hopefully, with this series you'll have a few more tricks though that you can use to improve your model! Good luck!
I will learn from him to (I have made a dumb decision to replace System Identification course with another one).
Brilliant, just as my exam is coming up!
If I don't really know what the order of the system might be, I would first try to measure the frequency response function of the system, for example using tfestimate in Matlab (not to be confused with tfest) which uses Welch's method to average out the influence of noise. From this FRF data it will be a lot easier (after a bit of practice) to guess what a order of a model would approximate the system well, at least from the data used to obtain the FRF.
I assume one of the next videos in this series might also cover this and also mention when the measured FRF can be "trusted" using the coherence (though that doesn't give a completely objective measure of the trustworthiness).
Best course ever !
Very clearly explained, I really liked it!
Great stuff Brian....still remember your spring and paint can analogy from back in the day.....one question though - this is something that I've seen controls engineer have misalignment on in citing some literature; when it comes to using System ID on closed loop system i.e. let say a PID based black box, this is common when it comes to using off-the-shelf components (motor drives etc), the method doesn't work well or is invalid in some people's view.....what's your take on using SysID in identifying a model of a system when a cascaded closed-loop is part of the system....
Very good, as usual with Brian
Great explanation!
Hello @MATLAB, Hi Brian, I'm trying to fit a model to my data which is a time series sports data. I have 2 inputs vectors of under 10 representative features each, corresponding to the 2 teams state vectors at time t-1, my output is either a win or a draw ( so 3 possible output ) so I would like to output the probabilities for each possible output. I feel dumb but I can't wrap my head around that .. Let's say A vs B was the game at t-1, and A won the game, next week the game is C vs A, my goal here is to take the state vector of A ( as a said earlier it would be a vector or few parameters that represents A ( Number of goals, shot on target, ranking at the time of the game ) PLUS the fact they won, and feed that to the next step ( which is the input of C vs Agame ) and so on .. up to 5 steps. Maybe you could give me some tips to get started ? Thanks and a big thank you for your Control series they are amazing
Great video but I have a question. In the example, how can you tell by examining the system response that you don't need to include any zeros in G(s)? I have learnt that the poles tell if the system is stable or unstable, smooth or oscillatory etc but I don't understand what role the zeros play. Thank you and keep up the good work!
First of all, I like this series about controltools in matlab . I have a question .....actually it may sound silly.
Why Mathworks has suddenly( not that suddenly...of course :D) started to make video tutorials about its tools although some of these tools were introduced more than 7 or 9 years ago?
I am not aware of the whole process of product or software release , I know to sure tutorials take time to prepare....but not to the point that the company produces a tutorial for some of the tools inside its software after 9 years and maybe even more.
14:52 The rigid cable is connected to mass m and its displacement x1 is the response signal; pulley 1 is rigidly fixed to a wall whereas pulley 2 is linked to the wall with spring k. The center of pulley 2 undergoes displacement x2=x1/2. The second additive term in the differential equation should be ... +k*x1/2 instead of ... +k*x1/4, is it not? Thanks in advance for the response.
hi.. for a step response test for sample a heater.. what is the input and output of the system? is it the pid controller out (as input) and temperature (as the output)?
Great stuff as usual and perfect timing! Could you please include information: 1. How to setup experiments to obtain good quality data and some common input (excitation signals) to use and caveats 2. MIMO state-space estimation in this series?
Hi Brian that was a great video but I have a question, where can I locate the block you use at minute 9:46? The one called real system with uknown dynamics, I can't find it and it would be useful for my college project.
Regards
That's just a regular Simulink subsystem that I labeled as "Real system with unknown dynamics". I created the dynamics in Simulink and then hid them away in a subsystem so the viewer couldn't see it. You can think of that block as whatever your real model or physical system is producing.
How can you know a mass and poulies system work from 1st principles .. and I didn't. What are the drawbacks of never using 1st principles to solve problems
God bless you
could you share the matlab code and simulink file please?
Often there is a delay (exp(-s*T)) in real systems.
Are you Brian or brain? ❤🔥
V GOOD!
I have thought that Brian Douglas is dead (he has not posted anything his channel from the period a SARS-Cov 2 vaccine was not discovered).
Hmm, I can't argue with that logic!