search

MarlinFirmware / Marlin

Marlin is an optimized firmware for RepRap 3D printers based on the Arduino platform. Many commercial 3D printers come with Marlin installed. Check with your vendor if you need source code for your specific machine.
https://marlinfw.org
GNU General Public License v3.0
16.19k stars 19.21k forks source link

[FR] Electrochemical Machining (ECM) and EDM Retraction #22849

Open JohnAtMorlock opened 3 years ago

JohnAtMorlock commented 3 years ago

Is your feature request related to a problem? Please describe.

Me and several others have been using Marlin for our ECM (electrochemical machining) CNC machines. ECM Allows us to cut metal without using any physical force. (https://youtu.be/KieJN-J4s38)

ECM is sort of like EDM, in that current must be kept constant by carefully controlling how far away our tooling is from the work. Usually, this is a few hundred microns. If the tool gets too close to the work, it shorts. If it is too far away, the current drops.

We need to run the printer backwards when the current gets too high or drops to zero (shorts)

Are you looking for hardware support?

Cheap current sensors report their values in the form of an analog value between 0 and 5 volts. All that is required is a way to declare a "safe" range of analog values

Example) Minimum_Sensor_Value = 1023; Maximum__Sensor_Value = 1023;

The value/current relationship changes from sensor to sensor

Describe the feature you want

We require a feature that detects when a sensor reports the current is too high or low. Similar to a filament runout sensor in theory, the behavior of the printer is controlled by an external stimuli.

Theory of operation: 1) A sensor pin receives an analog value from a current sensor 2) If the analog value is too high or too low, an event is triggered 3) The event is as follows: The machine reverses its last move. G1 Z0.05 F600 => G1 Z-0.05 F100 The reversal is continued until the sensor value is within the declared range. 4) The program resumes at the line of gcode it first started reversing at

Example) Move to point A Move to point B Move to point C Move to point D Move to point E DETECT CURRENT IS TOO HIGH Move to point D STILL DETECT CURRENT IS TOO HIGH Move to point C CURRENT RETURNED TO NORMAL Move to point D Move to point E Move to point F

Essentially, the machine needs to run its gcode backwards until the sensor reading goes back to normal. This would be conducted differently depending upon if the machine was in relative or absolute positioning mode. It would also need to search for not only G1 commands, but also E1 since we use the E-axis on our machines as a cheap and easy rotational axis.

Additional context

No response

DerAndere1 commented 3 years ago

Marlin now has support additional axes, see https://github.com/MarlinFirmware/MarlinDocumentation/pull/409/files or https://github.com/DerAndere1/Marlin/wiki, so you do not have to use extruders for the rotational axes and you can add endstops to them. That approach may be a little more robust long term. Can you point me to code that calculates the required rotational speed for the rotational axes (depending on the current position of the other axes) to opbtain constant feedrate?

JohnAtMorlock commented 3 years ago

We have not been calculating the required rotational speed. I just tell tell the rotational axis to move to a certain position using hand written gcode. It is all touch and go.

G1 E90 sort of thing

DerAndere1 commented 3 years ago

For others trying 5 axis CNC: If you do not have implemented inverse kinematics for your machine, make sure you reduce the speed for the rotational A axis to account for the workpiece "diameter" while the tool is in contact with the workpiece, otherwise the actual feedrate (surface speed) is too high.

DerM4209 commented 2 years ago

Hi, I'm currently working on an EDM machine and I would also need this feature for it. I have almost no coding experience, but I really need the feature and so I will work on it until I can get it to work :)

I found out, that the feature can be used with dynomotion's kmotion software. Here is a video from BAXEDM about this topic: https://youtu.be/_pebUGo2o-E Unfortunately he is not showing the C code, that he used for it.

Here is a thread from the dynomotion forum with some discussion about the C code for controlling the gap based on the voltage: https://www.dynomotion.com/forum/viewtopic.php?f=7&t=680

Maybe the voltage controlled motion mode could be turned on with a Mxx command and later turned off with another Mxx to get back to normal feedrate controlled motion mode.

Maybe the requested feature could be achieved by continuously saving the last position every few milliseconds, as long as there is no short. If a short occurs the execution of new gcode pauses and the machine moves to the saved position. After that the machine continues with the gcode until the next short.

I could somehow figure out how to use the code from G60/61 to save and load the current position in intervals. So, only voltage reading and testing left...

DerM4209 commented 2 years ago

While trying to figure out how Marlin works, I had an idea how it could be done without saving the position in intervals, what likely would cause some problems. Maybe it could be done by saving just two positions for G0/G1 commands (a starting position and a target position) and moving back and forth between them until the material got removed. To decide if the machine should move towards the target position, towards the starting position or hold the current position an analog voltage 0-5V could be applied to an analog pin and a range could be defined like for example:

The workflow could then be like:

The feature should then work like:

Doing so, the machine should only reach the target position if the path is free of shorts.

For adding some safety it could be a good thing to let the machine wait a moment after reaching the target position until the arc has cut enough clearance to use this position as safe position for the next gcode section. Like, if there is no longer material close enough to ignite an arc at the target position continue with the next gcode section.

It will likely take some more days/weeks until I can get something to work, but I would really like to hear some feedback whether my plan is going into the right direction for the requested feature.

@JohnAtMorlock What do you think?

Thanks in advance :)

DerAndere1 commented 2 years ago

Currently G1 first uses get_destinatio_from_command() to update the destinationvariable. Then it calls https://github.com/MarlinFirmware/Marlin/blob/72b99bf1ba24cb9124668b958039b32a164c68cd/Marlin/src/module/motion.cpp#L1239

For ECM, G1 would have to call a modified version of that function where the arc voltage has to be monitored. Voltages below a configurable threshold should limit the actual feedrate (which probably means a recalculation of the planned trajectory (see planner.cpp)). The polling of the arc voltage could be done similarly to endstops: Endstops are implemented in endstops.cpp and handled in https://github.com/MarlinFirmware/Marlin/blob/b9ee84e8dcc6ae81aba3d46e079ed26e1a308a49/Marlin/src/module/planner.cpp#L1708), the endstop polling is called from the temperatur ISR: https://github.com/MarlinFirmware/Marlin/blob/b9ee84e8dcc6ae81aba3d46e079ed26e1a308a49/Marlin/src/module/temperature.cpp#L3518

DerM4209 commented 2 years ago

@DerAndere1, thank you very much for your reply, I will look at the files you mentioned. For the case that a short occurs, do you think moving back towards the target position of the last G1 command would be a suitable approach or could there be a better way to do it?

DerAndere1 commented 2 years ago

I have no clue about ECM. But from your link to the source code for the control loop (https://www.dynomotion.com/forum/viewtopic.php?f=7&t=680#p1369) you can see that there are three conditions: 

if (Voltage lower than SC voltage with some safety limit) 
  go reverse
if (Voltage higher than Upper GAP window)
  go faster
if (Voltage higher than SC but lower than Down GAP window)
  go slower
DerM4209 commented 2 years ago

I got some code ready for testing, so that I could drill single holes with it on my EDM setup. I uploaded the changed files to my github profile, but the code is not finished, yet.

Here is some footage of it: https://youtu.be/BT2hdXJYGeE

At the moment, it is just for a single move and missing a gcode command to turn it on and off, but I will add that soon.

DerAndere1 commented 2 years ago

Congrats! I was able to find changes in gcode.h/cpp and temperature.h/cpp by searching for "EDM" and "edm".

DerM4209 commented 2 years ago

Thank you :) While working on the code, I noticed some strange behavior when using the get_cartesian_from_steppers() function to get the cartes variable. I used cartes to compare it with my target variable to check whether the machine has reached the target. My problem is: When I run the code I can drill in millimeter steps, but when I try to drill in sub - millimeter 0.10mm steps, start from a position with decimal places or enter multiple commands in fast succession, the cartes variable gets no longer updatet / the machine never reachs the target.

Do you know what could be the cause, or could it even be a bug in the get_cartesian_from_steppers() function?

Is there another way to check whether the machine has really reached the target?

DerAndere1 commented 2 years ago

I did not spend a lot time tinkering about this, and I guess I guess I missed something fundamental. But in my understanding, Marlins loop() will be called after initialization, and loop() calls idle(). idle() calls ecm_edm_mode(), which (as long as the target is not reached) calls move_to_target(), which then calls idle(). So as long as target is not reached, you have an "inner loop" there and I think none of the other tasks of idle() are executed during this time:

also, the call to get_cartesian_from_steppers(); in move_to_target() is outside of this "inner loop", so it will not get called. This may explain that the cartes is not updated. To fix it, in file ecm_edm_mode.h, try to move the call to idle(); to the end of the function body of move_to_target()

DerAndere1 commented 2 years ago

Also research how to check for equality of floating point numbers on computers. Marlin has a macro for this problem, so

if (cartes == target_position) {

should be replaced by 

if (NEAR(cartes, target_position)) {

That might get rid of the second bug with decimal starting positions

DerM4209 commented 2 years ago

Thank you for your help, I tried to change the code like you said, but the problem still exists. I also reduced the code to a minimum, to see whether it could be caused by something else, but nothing changed.

It is easy to reproduce, moving in XY in 10mm or 1mm steps is not a problem, but as soon as you try 0.1mm steps it will no longer reach the target. The Z axis seems to be even more problematic, because after just two 0.1mm steps the problem appears, whereas in XY slowly executing a few 0.1mm steps is possible.

DerM4209 commented 2 years ago

Yes! I found another way to check if the machine has reached the target by using: planner.has_blocks_queued()

Now everything works as planned :)

DerM4209 commented 2 years ago

Here is the code that I've got so far: https://github.com/DerM4209/Marlin_ECM_EDM_Mode Can someone help me improving it, so that it can be integrated into the Marlin Firmware?

Here is another video of it in use: https://youtu.be/hA2yRkNJqhs

Cooper-8pTX commented 2 years ago

@DerM4209 I also do ECM work and will try out your code in the coming weeks. I've found that marlin seems to have a lot of trouble executing really slow movements. I don't know if that will be a problem closed loop though.

DerM4209 commented 2 years ago

Hi @Cooper-8pTX , thank you for your interest. Have you tried the code, yet? I have almost finished the build of a Wire EDM machine, so that I can test it out for that. EDM Drilling / Sinking does work so far. Sometimes it happened, that when the machine performed very small movements, the electrode did not move far enough away from the workpiece to clear the short, so that the machine could not continue in the program. I solved that by bumping the electrode, but I think the code needs some changes to move further away than to the last position if the short can not be cleared by that.

Cooper-8pTX commented 2 years ago

I've not had a chance to work on the firmware yet myself, but may in the coming month or two. I've found that Marlin doesn't like to really do much when the feed rate or distance are set too low. That may be part of your issue.

DerM4209 commented 2 years ago

In the last days I could make some progress with my Wire EDM machine. https://youtu.be/JpsVs2Xm-9E https://hackaday.io/project/181551-wire-edm-machine I think integration of Wire EDM (and also ECM) would be really useful for Marlin Users, because machines like the one that I built do barely cost more than 3D printers and you can precisely cut thick metal parts with it. So you can make all your flat / extruded 2D parts out of metal instead of plastic and you can also make 3D parts out of multiple metal pieces stacked on top of each other. It's also not as loud as CNC milling and so it's suited for apartments and working at night.

DerM4209 commented 2 years ago

Hi @Cooper-8pTX , besides Wire EDM I also want to try out ECM. I saw your Valentine-ECM-Mod and want to build it and test it out with the code for closed loop movement.

DerM4209 commented 2 years ago

@Cooper-8pTX , I spent the day building an ECM machine based on your Valentine-ECM-Mod to test it out with closed loop control. It worked very well and I really like the way ECM works - It seems like there is not much that can go wrong (like wire break with EDM) and by using closed loop control, the feedrate should never be too fast or too slow. At my first try the cutting speed was very slow, but maybe I can improve it somehow in the future. Do you have an idea how to increase cutting speed?

Here is a video of the test: https://youtu.be/hf4My9hoJ5M

I think ECM has a lot of potential and I really want to learn more about it.

e7leef commented 1 year ago

can you contact please, im interest in your ecm machine video on youtube and i really want help on it m9839452@gmail.com @DerM4209

DerM4209 commented 1 year ago

@e7leef I've sent you an email.