Above, the final version of "Leftshark" Drill (so named after the Left Shark dance in Katy Perry's super bowl show :)...). Here an arduino and potentiometer controlled the frequency of the stochastic perturbation, and thereby able to characterize a wide range of materials while optimizing feeds and speeds to drill more efficiently into the material of interest.
TIME OF FLIGHT PROTOTYPE
We prototyped a time-of-flight approach: a laser beams light, reflects off a surface and returns back to a photodiode. With a high precision quartz crystal clock, we got the time delay and measured the distance the drill was from the surface.
ARDUINOS, DAQs, VOLTMETERS, SCOPES, AND MORE!
Left to right: (1) Computer with dithering code (random PWM of voltage input); (2) Scope connected to voltmeter and current to voltage converter, monitoring current and voltage; (3) The arduino; (4) The drill; (5) 80-20 rails with our material of interest bolted in and connected to a set of constant force springs sliding the material into the drill bit; (6) A webcam connected to MATLAB monitoring the speed of the advancing material.
A top view of the experimental setup.
A close up of just the drill, the 80-20 setup, and the webcam.
Connecting to the DAQ for data collection of voltage input and power output.
THE MATLAB PROGRAM
In order to monitor the speed of the advancing material, and see whether the stochastic voltage input actually increased drilling efficiency, I wrote a MATLAB code that followed the red-tape on the rail as it was advancing towards the drill. With position and time data, I was able to extrapolate velocity for each iteration of dithering code.
AND THE RESULTS ARE IN!
Above, we calculated the impulse responses for each material of interest using a Toeplitz matrix inversion method. Essentially, we had the stochastic binary voltage input, and then measured the power output of the drill. Using a special matrix called the Toeplitz matrix we were able to use deconvolution to determine the impulse response for each of the materials.
DETERMINING ACCURACY OF IMPULSE RESPONSES
In order to determine the accuracy of our impulse responses, we re-convolved the impulse response with the voltage input to see if we would recreate the power output. With over 96% accuracy of the convolution, we can safely say that our method had accurate predictive power.
An Ashby-chart, which inspired us in characterizing materials into different families based on the parameters of the drilling impulse response.
AND MOST IMPORTANTLY, MATERIAL CLASSIFICATION
A smart-drill could ideally identify materials from impulse response and then in real time, optimize the feeds and speeds of the drill bit to accommodate a more efficient drilling operation. For each of the impulse responses, our team calculated the Fourier transform and determined the natural frequency and gain of each; graphing them above and establishing clusters of materials with similar mechanical properties (likely related to hardness, density, ductility, etc) for drilling.