So the notion that in order for me to proceed scientifically (and practically) towards the design of rotor dynamics identification algorithms that will run in real-time, I first need to model the effects of the rotor failure through experimentation.
The two best ways of the doing this are (1) Buy a +R2k Tachometer and record (probably by recording the display) the speed of the propeller with a pre-determined PWM value. (2) Use the high speed capability of an action cam (such as GoPro) and some clever algorithms to compute the propeller RPM. The latter is the cheaper (and the geekier) option of the two.
The setup was such that contrast was created through the use of a black mat cloth on the setup table and painting the opposing blade of the propeller black and white (see below). A RGB (red-green-blue) adaptive algorithm was developed which would mitigate the occurrence of glares on the blade which would in turn give a false reading with the algorithm.
Each frame was analyzed by through creating a circular ring of data points around the center of the rotor (choosen through user input), then analyzing the RGB values distribution along the circumference in order to locate the position of the blade (see below). The generation of PWM commands was achieved through the use of an Arduino Uno script with allowed a 2 seconds delay between each command in order to generate enough data points for the estimation algorithm.
The computation of RPM was arrived (particularly at high RPM) through averaging the angular values for each frame (with the assumption that the blurring effect on on each subsequent frame is very similar) and then using the framerate (240 frames per second) to compute the revolutions per second. This was achieved and shown below.
The RPM measurements were then filtered using a Butterworth filter and a total variation diminishing algorithm which resulted in the results below. It does show that this method has great promise and the subsequent objective is then to model the dynamics using high order transfer functions (up to fourth order) for analysing the change in model parameters once the faults have been introduced. Your comments are welcomed!
The two best ways of the doing this are (1) Buy a +R2k Tachometer and record (probably by recording the display) the speed of the propeller with a pre-determined PWM value. (2) Use the high speed capability of an action cam (such as GoPro) and some clever algorithms to compute the propeller RPM. The latter is the cheaper (and the geekier) option of the two.
The setup was such that contrast was created through the use of a black mat cloth on the setup table and painting the opposing blade of the propeller black and white (see below). A RGB (red-green-blue) adaptive algorithm was developed which would mitigate the occurrence of glares on the blade which would in turn give a false reading with the algorithm.
Each frame was analyzed by through creating a circular ring of data points around the center of the rotor (choosen through user input), then analyzing the RGB values distribution along the circumference in order to locate the position of the blade (see below). The generation of PWM commands was achieved through the use of an Arduino Uno script with allowed a 2 seconds delay between each command in order to generate enough data points for the estimation algorithm.
The computation of RPM was arrived (particularly at high RPM) through averaging the angular values for each frame (with the assumption that the blurring effect on on each subsequent frame is very similar) and then using the framerate (240 frames per second) to compute the revolutions per second. This was achieved and shown below.
The RPM measurements were then filtered using a Butterworth filter and a total variation diminishing algorithm which resulted in the results below. It does show that this method has great promise and the subsequent objective is then to model the dynamics using high order transfer functions (up to fourth order) for analysing the change in model parameters once the faults have been introduced. Your comments are welcomed!
such a nice piece of article. Thank you for sharing your knowledge. This is a very valuable thing you shared. good stuff!!Licensed Drone Videographer Dubai
ReplyDelete