-
Notifications
You must be signed in to change notification settings - Fork 1
Progress Report on Data Measurements
Encountering bug with printer where printer believes heat bed is at 0* Centrigrade and heats up to specified number from Cura, even thought the heat bed is starting off at a non zero temperature (like 24* or 48*). This only happens if I set the temperature for the print bed in Cura. I cannot figure out how to pre-heat the print bed on the printer yet. However I have been able to print by manually heating up the print bed while the print is happening, so at least I know that the printer will function and it can change heat bed values on the fly.
I printed a 40mm cube which came out pretty well except for some obvious greater density at the bottom (potentially b/c print bed was unheated?). Also printed 10mm cube with large error forming convex shape. I hope to find the root of the this bug or at least set the default heat of the print bed to be 60.
Possible optimization: Wait for heat bed to warm up before heating up nozzle and beginning extrusion. Saw a couple posts about that online (https://ultimaker.com/en/community/4015-preheating-in-cura-first-heatbed-second-nozzle). This wouldn't really solve my problem though because it is about the printer being unaware of its own heat bed level. Which is so weird that Cura would affect that. Maybe there is a gCode function that can check the temperature of the heat bed and calling that would fix the issue.
I compared my Cura settings with a colleague and them to be the exact same. The print bed issue still happened however. My colleague might have been printing at a higher temperature than she thought because of the glitch. I also tried leveling the print bed again to no avail. I think what I am going to have to do is use a thermometer to measure the temperature the print bed starts at so that when I set the print bed settings in Cura I just know it is going to add onto that original temperature. A shame that this is such an issue but I see no way to do it otherwise. There also does no seem to be a way to preheat the printer.
Since my colleague has had success printing multiple things in a row I wonder if the printer will actually keep track of the change in temperature of the print bed for multiple prints. Because right now I can see an issue where if I heat the print bed to 100 degrees and move onto the next print the printer will start from 0 degrees and heat up the print bed beyond 100.
In other news all of the models I have been printing have not showed any signs of large cracks in their construction. This is unfortunate because it makes my whole experiment kind of useless. I need to think of a different parameter to measure and try to improve than just cracking. I am thinking a good kind of measure might be warping within the cubes. I noticed in the small cubes that the print bed temperature seemed to affect the warping seen in the walls of the cube, if i can find a temperature that is ideal for a 10mm cube, then a 20mm cube and a 30mm cube I might be able to derive some type of relation between size of cube and ideal bed temperature heat.
Today I am going to attempt to hack the gCode of my Cura stl files to heat up the print bed to the correct temperature before the print begins. Using the gCode setting M109 I tried to get the heat bed temperature, but that did not appear to do anything. Plus that command might just send a json file to some output and not even have anything to do with setting the current temp of the heat bed. It turns out that WANHAO has its own version of Cura which I installed and used, it did change the gCode a bit but did not solve the temperature issue. My original observation that the printer correctly estimates the print bed temperature if it is not giving an original print bed temp was actually incorrect, because it was just adjusting the extruder temperature which it has a working sensor for. It seems to be that the sensor for the heat of the print bed is broken in some capacity. Also the print bed appears unleveled, for when it started to print the PLA did not adhear and I had to manually heat up the nozel and print bed to scrap off the scraps. Hopefully I can just pre-heat the print bed to the correct temp which I judge with a thermometer, and use that as the print bed temperature for my experiment.
According to help from the WANHAO facebook group the issue is with something called the thermistor which I should replace with a "100K thermistor". I am quite unfamiliar with the hardware make up of the D6 so unfortunately I do not know what to buy and where. Hopefully someone gets back to me on the FB page but so far googling has not given me that much help either in determining what type of thermistor to get. Since this issue does not look like it is going away any time soon, and since the prints of the 10mm cube without any heat provided to the bed seem to be coming out fine, I have decided to draft an experiment that takes the heat bed to be around 28-35 degrees and only adjust the extruder heat.
Its time to start the experiment. I will first create the 12 different gCode exports with the temperature ranges, then input the data into a google sheet also recording the temperature of the heat bed at the time (which apparently the printer cannot do without a new welded thermistor 👎 I was linked an interesting paper by my colleague: https://files.slack.com/files-pri/T0EJFTLJG-F1NN2KYVC/paper5.pdf, I looked up the Taguchi method because I am unfamiliar with the method and most of statistics in general. From what I garnered, the Taguchi method involves maximizing the function
s/N = = 10* log_10(u^2/sigma^2) where s is the standard deviation of some function f, u is something called the "power of signal factors" and sigma is the "power of noise factors s"
I ordered a book on the Taguchi method and will read it, maybe using it for another experiment. In other news, I updated the experiment method to describe how I will measure each side of the 10mm model. To measure each model side I will draw a 10mm x 10mm square on a sheet of paper, line up one edge of the side of the 10mm cube and measure the offset on the other 3 edges. I will then use the max of these 3 offsets to get the distance off model each for each side.
The first three prints were complete failures (plastic strewn everywhere, clogged extruder ect.), meaning that I need to do a bit more testing to see what extruder temperature will actually cause PLA stick to a 30* C buildplate at all. However I did find that using a brim with the default settings of 20 lines did help with adhesion a little bit, and so did heating up the build plate. However since heating up the build plate is such a hacky process (because it is either heating up or cooling down with my broken thermistor) it would be better to just test extruder head temperatures that give something that even remotely resembles a cube. Thus I am going to change the settings on all my G-code outputs to use a platform adhesion type Brim with a size of 20 lines.
The results of my probing are as follows
160 * C Raft - Failure
180 * C Raft - Failure
185 * C Raft - Failure
190 * C Raft - Success (Tried it two times)
Interesting that 190 worked both times I tried, a difference of a mere 5 * C seems to be the difference between complete failure and pretty decent success. Also- I am not even heating the build plate! Why have a heatable build plate when success is about the same either way? This might imply an optimal difference between build plate and extruder temp to promote just enough adhesion. Ex. if your build plate heat sensor is broken, up the extruder temp.
Today I am going to find the upper limit for a successful print of the 10mm cube with no heated bed. I also need to choose whether to use brim or raft to promote adhesion.
160 * C Raft - Failure
180 * C Raft - Failure
185 * C Raft - Failure
190 * C Raft - Success (Tried it two times)
220 * C Raft - Success (But raft was hard to remove)
255 * C (max extruder heat on printer) * C Raft - Success (But raft was very hard to remove)
220 * C Brim - 1 Failure 1 Success (brim was heavily warped in either case)
255 * C Brim - Success (brim was hard to remove)
Given the fact that I am printing an object that can be printed so quickly, there is not much of a print time difference between a brim print and a raft print. In addition, brim prints seem to be more prone to failure by my limited testing. So I have decided to use raft's for this experiment.
One thing I noticed was while doing the experiments I noticed the extruder would extrude when it was doing its fast initial motion to break off the excess PLA from the extruder priming stage. This doesn't really make sense and only seemed to contribute to errors within' the model.
Since successes range from 190-255, I have decided to choose this range for my data measurements.
The raft and the brim would be so hard to remove on many of the models that it was hard to determine what was a "successful" print, after all, the cube might turn out perfect but the bottom would be subsumed by a tough raft that was so hard to remove that the bottom would have a large visible error.
Since there is an option for no adhesion type, a.k.a no raft or brim, in Cura, I have decided to try this out to see if there are any successes. It may be the case that an optimization of a print with no adhesion will still benefit prints with rafts or brims. I created 19 G-Code files with no adhesion on the 10mm and temps ranging from 160-255 * C in steps of 5. (Couldn't go up to 260 * C b/c of printer).
The results of these test prints are below:
160 * C No adhesion - Failure
190 * C No adhesion - Failure
200 * C No adhesion - Failure
205 * C No adhesion - Failure
210 * C No adhesiom - Success
(kind of, messed up bottom but ill just measure that), 2nd attempt worked perfectly
220 * C No adhesion - Success
255 * C No adhesion - Success
Given the fact that 210-255 * C wont completely fail, I will use these with no adhesion for my experiment. I will start collecting cubes in labeled boxes.
Today I began printing the 210 * C No adhesion 10mm cubes. Most of them were successes, but a few were complete failures once the bottom of the print came lose. A lot of these failures were very noticable when it came to correctness, so I've decided to leave them in the measurements b/c they should come out as terrible measurements. However there was one failure were the top couple layers simply failed to print and extruder got stuck on the top of the cube extruding into the 25% infill. This will be a bit harder to classify because all the sides look pretty correct, but the print itself was a massive failure. Perhaps I should establish a rule that if a print has "failed" it gets a 0 for its measure of accuracy no matter what.
Nothing much to report. Finished the 215 * C 10mm cubes. I was thinking it might be an interesting idea to model the optimal settings as simply a probability that a model will fail, given that the failures of the cube are so extreme. Also, having a simple "fail" state will be hypothesizing a tolerance for an acceptable cube and also allow me to categorize the extreme failures that result (otherwise I guess I would just measure the failure and get an outlier measurement that would bleed into my data). I will think about this decision more tomorrow. After all, I have a lot of time to print before I start taking measurements, given the rate at which these 10mm cubes are churning out.
Printing out 225* C cubes
Fixed the print bed temperature issue by removing the thermistor from the print bed via soldering and then soddering in a new thermistor. The printer now correctly heats up. I have created a Taguchi OA to begin an optimization experiment on many parameters that I will write about in the experiment setup portion of the wiki. I also downloaded Simplify3D to slice and get an idea of what I want the levels of my parameters to be for the experiment. Overall it looks like the experiment will have around 100 trails, which is the same amount as the 10mm cube experiment.
Today I am going to try and determine by how much each of my parameters should be adjusted from the default value to form the 5 levels. I am going to try and just look up recommended heuristics online.
On a slightly unrelated note, an interesting idea for a very "dumbed down" slicer for use by the average joe. Could use 3 sliders, "speed, strength, quality" that the user could adjust accordingly and then the parameters of the print would be individually adjusted without the user having to go in and specifically change those values. Would make desktop 3d printing way easier I would think and I have yet to see a slicer that does such a thing.
After determining the levels of my parameters. It is time to start the first data measurements. I printed out a test sample of the model and found my digital calipers were inadequate to measure it, also they were out of battery. However if I layed the model on a piece of paper and marked traces the cirlce I was trying to measure via pen I could then measure the error from the markings on the piece of paper. The pen I used to measure these markings has about a 0.5 mm width. I will measure the pen marking from the farthest edge to correct for this.
Hopefully I can get my hands on a micro-engraver or a very thin tipped pen to reduce this problem. Now all that is left is to replace the PLA and print bed surface on the printer to start the prints fresh.