Vibration sensor

[Horsedorf]

Originally Posted by extrapilot - 1. I think you will find that this approach is very similar in concept to knock detection in modern automobiles/ecus. That is, you are looking for a vibration signature that is buried within a lot of machine noise.Rob

Wellll.. yes and no, right? I mean an engine knock sensor is a microphone that is not axis specific. While an accelerometer *IS* axis specific, so if you mount it to the spindle collar you can already guarantee that the vibrations you are detecting along the axis of interest is generated by the cutter.

Though the last time I did anything like this was over twenty years ago and it was only peripherally in that it was a side project in a company I worked for. I only listened into conversations about it. however, they used a dual 68000 based system running at 10mhz to implement it. I left before it was done so i don't know if they succeeded or not.

[efrag]

thanks everyone for your input, I think I'm going to turn on PICs. Extrapilot thanks for the idea on "knock detection" I'm looking for it now.

[apache405]

Horsedorf is right, some signal conditioning will help reduce the load on the stamp's program. maybe use some logic gates and an RC circuit tuned to the desired frequency that is detected. I think if you use a two channel AND gate with the accelerometer and the RC on the inputs, when the RC circuit and the vibration are at the same frequency the gate will open. That will create a high signal that the stamp could process. Then in software there could be a counter tied to a timer so that in effect you have a number of high outputs per second limit that would then trigger a high output to your controller or the keyboard interface.

[extrapilot]

HD-

I dont think axis is really a factor in this. That is, this is a coupled system- would chatter always be in the same axis? What if you were cutting Y vs X, assuming a cut with the same 'side' of the cutter (now 90deg out of phase). Or, on a 45deg?

On the PIC stuff, this sort of work is not trivial. If all you are looking for is a peak acceleration, this is really simple- just use a nice opamp with the accel chip and charge a small cap in parallel with the ADC input on the PIC. But, I dont think that you can just say 'if accel voltage on X or Y axis > 2.64V then 'chatter detected.'' I hope Im wrong here and just a simple peak voltage detector alarm will work- and it might for a particular cutter/table geometry/stock combination. But I think you want a generic solution for different cutters (and flute counts), machine positions, and spindle speeds, which in combination implies substantially different chatter frequencies. I dont think you can count on any static RC filtering approach if the fingerprint can have 1+ order of magnitude difference in frequency. You are probably going to need to sample the noise as a large number of events and then use FFT or some other filtering approach to characterize it

Regards,

[apache405]

The axis of motion is critical in acceleration because it is a vector. A subroutine in the software could be used to determine the magnitude of the vector using the Pythagorean theorem in 3D. That way it detects vibration in the three axises of motion of the spindle.
Oh and who said the RC circuit had to be fixed? With a digital pot, the circuit can be tuned in software to match the material/cutter.

[extrapilot]

Apache

Im fairly certain that the line of inquiry here is 'is it feasible to use an accelerometer to detect chatter- and adjust Mach3's feed rate in a closed loop, without a lot of complexity?' So if that is correct, my perspective on this is- no, and probably not without dsp regardless. If one has to predefine resonance points per cutter/mill geometry, map that into a uC to control a digital pot, and then manually or digitally define the condition of the mill (cutter/spindle speed/stock/etc) for each change of configuration so that the uC can control the RC circuit tuning- that does not seem to me very practical- even if the chatter vibration is fully transmitted to the spindle vs. dissipated in the tool or stock/table (that is- the vibration at the accel chip is substantially above background noise).

Again, I could be horribly wrong in my assumption here- maybe chatter is always the same frequency regardless of config. That has not been my experience, however. I look forward to hearing the results of some tests- if it does work, it would be a nice failsafe

Best,

[apache405]

Extrapilot:

Point taken.

[IQChallenged]

Originally Posted by efrag Hi all, I'm building a vibration sensor for mills and Lathes. I'm using a basic stamp

I'm sure I'm the only one who doesn't know. But what's a "stamp"?

[Horsedorf]

Originally Posted by IQChallenged I'm sure I'm the only one who doesn't know. But what's a "stamp"?

They are a small , dip package that runs basic. They are easy to program using a windows integrated development environement and the pins can be assigned as inputs or outputs. They know how to do things like I2c and a/d conversion and etc. The major problem with them is that they are relatively slow.