In what way does geometry affect build rate in power bed fusion?

Hello!

I hope this is the right place to ask this question.

In power bed fusion, there are two main factors that drives the total build time on a part, these things are:

♦ The recoat time - Which is the time elapsed between the point when laser stops and the point when laser starts again, when a new layer of powder is applied.
♦ Build rate - The volume rate in which material kan be melted by the laser.

The recoat time is very easily calculated since every layer has the same recoat time. The build rate however, is something entierly different. The difference in build rate can be substantial when comparing different geometries.

Look at this video: [www.youtube.com]
In the video you can see that the laser "beam" seams really wide, and it seams to moves at a constant speed. This would indicate that a high build rate can be achieved only when a large portion of the potential width of the beam is utilized. In other words, geometries with really thin features that only utilizes a small portion of the beams potential width, requires a lower build rate.

What do you folks think about this?
I'm grateful for ALL the information you might provide me with on the subject!

Some questions:
What is a normal maximum width of the laser beam? For example on the EOS M290
What kind of geometries would require a low build rates?
What kind of geometries would require a high build rates?
Is there any way of calculating (or estimating with good accuracy) the average build rate for a part, before the manufacturing, without using a software that analyzes the 3D file?

Edited 1 time(s). Last edit at 06/09/2015 12:42PM by nesirglana.

What you see as the "beam width" is actually the laser moving back and forth very rapidly within that strip. Why it takes multiple passes instead of scanning the entire layer as a single pass I don't know. It may be due to the design of that printer. From the tech specs of that printer, beam has a focus diameter of 100 μm (0.004 in) and a scan speed of up to 7.0 m/s (23 ft./sec). Using a focus diameter of 100um, in 1 second it could fill in an area up to 7m x 100μm, or 700mm^2.

You'd have to know the total area of all layers being printed, or a total volume of the entire print to compute how long it would take to sinter the material with a best-case scenario of no overscanning, no travel time, and optimal material properties. You'd have to add in the recoat time for each layer times the number of layers to be printed. How you would figure out the area/volume to be printed without looking at the model is going to be difficult for anything but the most primitive of shapes. But you can probably approximate it by using an overall bounding box and get somewhat close.

It's also going to depend if the object is being printed out as a vector or a raster image. Imagine you are printing out a section of thinwall tube. You want it only 100μm thick. It could theoretically print with just a single pass of the laser just the width of the beam so it could print extremely fast per layer if the beam only has to trace the diameter of the tube. If it has to scan back and forth the entire area of the tube, there is much more area that would need to be covered even though it's not sintering anything in that area so would be much slower. However it would be easier to compute paths as you're just scanning the model not tracing it.

Quote
cdru
What you see as the "beam width" is actually the laser moving back and forth very rapidly within that strip. Why it takes multiple passes instead of scanning the entire layer as a single pass I don't know. It may be due to the design of that printer. From the tech specs of that printer, beam has a focus diameter of 100 μm (0.004 in) and a scan speed of up to 7.0 m/s (23 ft./sec). Using a focus diameter of 100um, in 1 second it could fill in an area up to 7m x 100μm, or 700mm^2.

You'd have to know the total area of all layers being printed, or a total volume of the entire print to compute how long it would take to sinter the material with a best-case scenario of no overscanning, no travel time, and optimal material properties. You'd have to add in the recoat time for each layer times the number of layers to be printed. How you would figure out the area/volume to be printed without looking at the model is going to be difficult for anything but the most primitive of shapes. But you can probably approximate it by using an overall bounding box and get somewhat close.

It's also going to depend if the object is being printed out as a vector or a raster image. Imagine you are printing out a section of thinwall tube. You want it only 100μm thick. It could theoretically print with just a single pass of the laser just the width of the beam so it could print extremely fast per layer if the beam only has to trace the diameter of the tube. If it has to scan back and forth the entire area of the tube, there is much more area that would need to be covered even though it's not sintering anything in that area so would be much slower. However it would be easier to compute paths as you're just scanning the model not tracing it.

Thank you for your answer cdru!

When i said "without using a software that analyzes the 3D file", I meant the slicing software that calculates the machine time! I didn't mean that you couldn't use CAD too check the volume of the part.

You say that you could calculate the build time if you knew the volume. I would disagree, I've studied alot of parts that have been printed in an EOS M290. Since the total recoat time is easily calculated for a part, that time can easily be subtracted from the total machine time to get the average build rate for that part. In the parts which i have studied, the build rate varies ALOT, between different geometries. So I don't think the volume and the height (for recoat time) is enough to get a good estimation for the total machine time. I think you would need a software that calculates "tool path" for each indivdual layer.

Edited 1 time(s). Last edit at 06/09/2015 06:17PM by nesirglana.