Page MenuHomeAleph Objects Inc

Dual Testing
Closed, WontfixPublic


Testing needed on the new Dual

Related Objects

Mentioned Here
T1: Test task

Event Timeline

DaniAO created this task.Feb 6 2018, 11:20 AM
DaniAO triaged this task as Normal priority.
DaniAO created this object with edit policy "LulzBot Hardware Products (Project)".
nickp assigned this task to tutley.May 8 2018, 9:20 AM
nickp added a subscriber: nickp.

Update for testing.

Mark is working on the board for the toolhead interface now.
Marcio mentioned in today's meeting (5/10/18) that there is firmware to test with a single. Firmware for the dual is being worked on.

nickp raised the priority of this task from Normal to Needs Triage.May 10 2018, 11:06 AM

A simple toolhead interface board has been constructed and and tested on both toolhead designs. Also, both an Arduino Uno and Arduino Due have been set up and tested to drive the toohead servos. The input plug of the toolhead interface has been modified to attache to both the Uno and the Due.

As I was able to program a Due to operate the servos, I will now attempt to write test code for the Archim board to drive the servos directly. Until then, either the Uno or the Due can be used as a bridge between the Archim and the toolhead servos.

west added a subscriber: west.EditedOct 2 2018, 2:30 PM

4 machines are running

We had an actuator fail on the longevity test after being cycled 140k times. This equates to 914 dual benchy prints completed. I have contacted actuonix about the failure to see if the actuator lifetime can be improved beyond this.
At this rate, this would be 2.5 benchy prints every day for a year before failure.

Is this lifetime sufficient?
Personally i believe so, considering the lifetime and cost of other commonly replaced items. The heat bed is the first to come to mind. RAMBO, motors, pulleys, toolhead components, etc.

The actuators are easy to replace, and cost ~60$ ea direct from actuonix. So a customer could pretty easily replace them for ~120$

I would like to hear others' opinions on the matter.

karrad added a subscriber: karrad.Oct 16 2018, 9:49 AM

I am a little concerned about that. While 2.5 benchies a day for a year doesn't sound too short, how many larger prints would that be in a day? It seems like something 2.5 times as tall with similar geometry would only be one print a day. Going to something that is more than 3 times as tall may cause the failures before the year mark at 1 print a day. With the build volume capabilities of the TAZ, I am assuming we will have a large percentage of customers in that niche.

kent added a subscriber: kent.Oct 16 2018, 9:52 AM

@tutley were you able to determine the cause of the failure? I'm wondering if it was the wires breaking or the the brushes wearing out or some other mechanical failure. If we can identify the weak link and make it more robust, we might be able to increase the MTBF.

@karrad this wasnt based on the height of the model and number of layers, it was based on the number of toolchanges in the print. So something may be 3x as tall but the number of toolchanges isnt necessarily 3x as many

@kent it was a mechanical failure as far as i can tell, there are little metal tabs for a resistance measurement that got bent and gouged the resistive pad.

@tutley Can we get any data on the average number of tool changes per print, and perhaps different types of dual prints? Support, dual color, etc. The reason I decided to focus on height is due to more layers. in my mind, more layers = more tool head changes on dual prints.

@karrad sure i can look into that a bit.

FWIW The full build volume for quiver is 290mm in z and if you printed at .25mm layers, you would have 1160 layers. If this was a dual print with a toolchange on every layer, you would get 236.5 full build volume prints before failure.

from actuonix " At 140K cycles, failure was probably EOL from the motor brushes. Those contacts are the same ones we have been using for years in our PQ12 series and we have had no other issues with those.
140K cycles we would consider very good lifetime, I don't know if we will be able to coax more life out of them."

This is also dependent on how many toolhead changes are present in each print. This doesn't mean anything to a single-extruder print, but is more of an issue for a print that is dual-material/color for every layer.

Using the benchy metric, this would come out to 5h15m a day for a year. The benchy is a small but moderate tool change part - not all layers have a tool change, but most do.

Going off of the 140k number and not the benchy conversion metric, assuming a full build height print using the 0.18mm layer height PLA profile, this would come out to 174prints. I created a 290mm tall 10mm square tower, which I call "worlds_dumbest_dualprint" to get a metric for time based on a worst case scenario. worlds_dumbest_dualprint was sliced without a prime tower at 0.18 height on the polylite PLA profile and it came to 7h34m in Cura Estimate Time. Again, this is an extreme case. If you print 2 a day, this results in 87 days - roughly 3 months. This would mean it would be at worst a quarterly maintenance. To test the other extreme, I made 2 towers of 270*135*285 to try and get close to max build volume. This print, dubbed "worlds_most_unnecessary_dualprint", was sliced with the same profile and came to 16d19h43m. if you run that back-to-back for 174 prints you get 2928.42days, or 8 years.

So, somewhere between 3 months and 8 years of continuous use depending on what you are doing.

UPDATE: It turns out our initial estimate for the number of toolchanges in the benchy print was off. Originally we just searched the dual benchy g code for the number of T0 and number of T1 commands called out to determine the number of toolchanges (~150 ea), however there are T0 and T1 commands associated to temping up and down the nozzles. So originally we thought there were 300 toolchanges total in the benchy print, it turns out there are acutally about 120 total, or 60 cycles per actuator. This means instead of simulating 914 benchy's before failure, we actually simulated ~2300 benchy's. or roughly 6.3 benchy's a day for a year.

I feel much more confident about the longevity of the toolheads.

This engine print with support has about 360 total toolchanges or 180 cycles per actuator. This means we would be able to print ~777 of these before an actuator fails. This is an 11 hr print

From actuonix "We have tried longer life brushed motors, but their lifetime was really no better than what we are already using.
Using a brushless motor would require a new PCB (brushless driver) and would be a different size requiring us to make a different housing to fit. This would require NRE fees, mold fees, 1000+ MOQs plus there would be a significant price increase as brushless motors are always more expensive."

The impossible gears designed for the V3 dual, when sliced with supports on 0.25 layer height, has 28 cycles. That comes to 5,000 prints of that model.

Since the motor brush lifetime is dependent on number of revolutions, and we are using the 100:1 actuator, we can likely lengthen the actuator lifetime by choosing a different gear ratio. Actuonix has a 30:1, 63:1 and a 100:1 option. The 30:1 is good for moving 15N (3.37lbs) , the 63:1 is good for 30N(6.74lbs), and the 100:1 is good for 40N (8.99lbs). We are only moving a mass of 0.85lbs.

My thoughts for choosing the 100:1 is that it will be very hard to backdrive so it likely wont lose position on a hard probe/failed probe or running over part of a print.

I think the 63:1 will work great, but i would be a little hesitant about the 30:1.

this large motor block print with supports takes 2days 8hrs and has 380 total toolchanges, meaning you could get 737 prints before a failure

@tutley After assessing this mornings actuator situation could you update us here with your findings, please.

Thank you

logan added a subscriber: logan.Nov 13 2018, 7:52 AM

@nickp Still testing a second 100:1, but here are the actuator results

Steven closed this task as Wontfix.Nov 16 2018, 3:43 PM

Closing this as actionable items are being tracked elsewhere.