Cast Knife results

Finally, the moment of truth has arrived. Did the mold lock design work? Are the two halves of the mold aligned, are the molds resting in place or held apart enough to generate significant flash? In this case, these molds were not even glued. Instead we weighted them with a steel block and attached a pouring cup after using a 3/4″ drill to make a hole to the parting line. The pouring is captured on video below for your review.

The pour went well, the material you see is a eutectic iron which is not the material of choice for making knives. However, this material is very fluid and can fill many details. If our mold halves are misaligned or likely to have significant gaps, we should see it with this pour. Shakeout was very satisfying and pictured below. The metal took exactly the shape we hoped for.

I wanted to use these to demonstrate how the final steel knife could look so I made a handle using parachute cord and wrapping the handle. There were a few areas were the mold left more roughness than desired. Therefore, we will play around with a mold coating before pouring the steel version. Anyway, I’m content with the process documented here and I’ll debating the next pattern for us to make in the lab. Please let me know you comments and thoughts about this process.

Knife pattern boxes

Using some 2×4 pieces, I completed the pattern boxes for the knife patterns that we made on the desktop cnc. This is good because it gives me a chance to mention wood sealers and my experience so far trying various paints. First, let me direct you to a common source for these supplies (at some point I hope to have a longer list of suppliers) at Freeman Supply. They have a nice catalog that you can search and find all sorts of supplies which also indicate the commonly used products in making metal castings. So, the top choices that are possible are Lacquers, Epoxies, Urethane, and Shellac. If you are finishing a floor or furniture there are some nice articles that discuss the differences between these products. The coatings have different amounts of protection, shelf life, ease to sandpaper, tolerance to heat, etc. However, we have one additional twist to the average need to protect the wood surface. In our lab, we use a phenolic urethane coldbox binder system. So, we can use any of the coatings but particularly the urethane tends be more difficult to protect and return to the original finish. For now, we are using a clear Lacquer to coat the patterns and we’ve seen pretty good results on other patterns.  The picture of the boxes that I made and how they look after the first coat of Lacquer is below.

Our next step is to pull a couple molds from these pattern boxes to see if the geometry fits the way we expect. If everything goes well, we will try to cut a temporary gating system and pour the next time we are melting down (likely friday, cast iron). Making these knives of cast iron is not the correct final material but we pour cast iron often and it’s a good material for checking out the geometry.

Image analysis casting manufacturing

We’ve found that image analysis techniques are interesting for developing manufacturing routes for castings. William Warriner is developing all of the routines shown below to be used in his PhD work. For example, watershed segmentation of the solidification profile of a casting geometry is quite illustrative. Below, at left is a component, shown transparent, and at right are the watershed segments of the component’s solidification profile overlaid on the component. Essentially the segments tell us what regions of the casting can be fed by the same feeder or group of feeders. Feeding will occur from segment boundaries where solidification begins and proceeds the thickest section.

Using attributes of each segment and the solidification profile, we can apply known feeder design guidelines to generate feeder geometries. Below are the feeders generated from the segments. In practice, feeders that interfere with geometry can be replaced by side-attachment feeders, or by gating directly to the location of the feeder. In the meantime, it is still worth visually reporting that a feeder is required.

Several of the sections are close and it is believed that number of feeders is correlated with increased cost and with decreased ease of manufacturing. One goal could be to decrease the number of feeders. One way to do so is to provide connections, or feed pads, between segments. Creating feed pads can be accomplished by drawing a solid tube between the feeder locations of neighboring segments. Below at left are the connecting tubes overlaid on the component. At right is the same image with greater visibility of the tubes.

One strategy for reducing the number of feeders is to cluster their segments by proximity. From the perspective of connector tubes, that would mean removing longer tubes and retaining shorter ones. Below at left are the original set of tubes, and at right are the reduced set. Note that currently there isn’t an obvious quantitative metric for doing this without human intervention. The tube removal process here is based purely on human intuition and is intended to illustrate the usefulness of the tubes.

Now that the connector tubes have been pared down to several clusters, it should be possible to reduce the number of feeders so that each cluster is fed by only one feeder. An example, purely for illustration purposes, is shown below. The number of feeders has decreased from 16 to 6. Note that the connector tubes are quite crude, and would almost certainly not be able to be implemented as they appear here.

A casting designer would have to work closely with a product designer to rework the original geometry. They would likely incorporate the connector tubes as wall thickness changes in appropriate locations. Any changes would have to avoid altering features that must remain as-cast. There is also the consideration of avoiding more net-shape machining by adding material in locations where machining is not required.

Presentations

I want to share a story from my graduate school days. So, I used to get really excited when something finally worked and I had a couple friends that I would rush to with the latest discovery to share it with them. One of those friends, Justin Garvin, brought something to my attention once that was really helpful for me later in life. Let me show you what I mean, take a look at the chart below. See how cool it is? Wiggles, flat spots, something is related to something, they even cross at some point… Man, there is a lot going on!

But, you can probably guess why my buddy, Justin, was laughing at me. No matter how insightful I am, or careful in the analysis and work… Because no matter how much work or painstaking care I put into generating the data in this plot, it is garbage without axis titles or a label or just about any details to describe what in the world I was thinking!

Seriously, it may seem trivial but these days I’m extremely focused on the details of plots, figures, and background information to make sure that we haven’t made a mistake in the interpretation of any result we make. These days especially given the data at our fingertips, it is even easier to draw conclusions from completely unrelated information just because they happen to be produced in the same sequence or at the same time. Take a look at Tyler Vigen’s site if you need some practical examples.

So, I’ve mentioned practicing presentations in an earlier post. In fact, I have more research to do (thanks, Alex Monroe) on the topic of assertion-evidence presentations. Let me offer a couple suggestions about preparing for this because the little I know about A-E is in support of this. I’d use the same approach to drafts of reports and proposals as well.

Your presentations need to tell a message, so ask yourself when you prepare it whether it communicates the message you intended to present. If it can’t answer a straightforward question, then it’s probably wasting a lot of effort and time. But that is in the end. In preparation, I will usually dump more information into my presentation before trimming back to the relevant stuff. Besides, you should know your audience enough to agree on a common language to discuss your topic. So the boring introduction stuff can sometimes be extremely valuable when you don’t even realize that the introduction is where the disagreement or discovery is. So consider that when you make your presentation. I hope to have more examples soon to post.

First pattern using desktop CNC

This is just a quick update on the pattern I started this morning after the activities yesterday. I made several adjustments to the parameters of the gcode path, including reducing the layer height and the stepover to reduce the drag on the router which I’ve decided is underpowered (300W). Those adjustments increased the machining time from about 4 hours up to 9 hours. So, this finished about 6pm which means the estimated was about right from my start time of about 9am. I’m happy with the finish quality but I’m not going to use this pattern except to explain the process due to the first mistake in the pattern of the knife on the right. About midway in the handle you can see where the wood was scarred by an operator error (I was trying to show how the machine could return to the G54 zero and the tool ended up dragging through the surface). If this hadn’t occurred then I would have used the saw to trim the ends and then build the box for the pattern.

So, next step is to cut this pattern again. I’m debating additional modifications to the gcode to speed up the roughing cut and using an angled cutter to add some natural draft to the edges. I’m pleased with the mold lock and the overall quality of the cut so far, keeping my fingers crossed about the quality of mold that we get from this.

Suggestions from speeding up so far include: cutting a border first to the depth slowly and then speeding up the internal roughing cuts, choosing a bigger router, better end mill.

Gcode for Knife pattern

So a little primer on how to go from STL file to gcode to run the desktop cnc mill. First, make sure you are absolutely happy with the STL file. If you export the file with too few facets then your machine paths will also be blocky. So this really is one of those circumstances where what you see is what you get. Also, I like to go ahead and reassemble the stl files together into it’s own model to check that the alignment is good, also that the coordinate system is in the right place and generally things are going to work out the way I like it. The picture shows that I’ve placed the origin at the lower left of the pattern and I’m showing the cope on top and the drag underneath. Maybe you can see that the knife sections line up and the corner of the mold lock matches as well. 

Next import these into your preferred gcode tool. Let me take a short aside on that topic because these are not as simple as the wealth of information on plastic 3D printing. In 3D printing looking at the free software world alone you can choose from Repetier Host, Cura, Slic3r, Skeinforge, just to name a few. These support gcode generation for many different 3D printer flavors and have many options for support materials, overhangs, and infill. I highly recommend looking at all of these if you are interested in 3D printing. Confusingly for experts in 3D printing, the cnc world breaks down into 2D, 2.5D, and for lack of better word 2.5D+ which would include the 3 axis mill and beyond. If you are just getting started with desktop CNC milling then I would suggest looking specifically at 2D such as jscut, FlatCAM, and PyCAM. Remember those tools are only going to cut completely through material, so they can be used for the mill, plasma cutter, laser, water jet, etc. For patternmaking, I’m interested in 2.5D machining which basically exercises the same muscles as the 3D plastic printing changing profiles from layer to layer in the height. It is similar to the 3D printing but instead of building up from 0 to the full height; it is subtractive taking from the full height down to the bottom.

For my 2.5D approach to the desktop cnc mill, I am using several tools because I haven’t settled on the best approach to teach with. First, Kiri:Moto is kind of amazing. If you even have a little interest in this stuff, you should check it out. Easy to set up, shows the machine paths, lots of default values that make sense, even integrated directly into Onshape. My experience with Kiri:Moto has been good, works quickly, and is useful for the initial check if you are going in the right direction. I have high hopes that this tool continues to get some love and attention because it is awesome! As is my style, I have broken it as well both when adding in new complicated stl files as well as on the export side. Unfortunately, there is not a lot of easy to debug log information about why the slicing or export fails. So, there are other alternatives… HeeksCNC, FreeMill, and for the low budget (but not free) hobbyist MeshCAM. I went ahead and bought a license of MeshCAM because the other alternatives I mentioned above are only Windows based, awkward to install, or a little gimmicky (as in not clear what the terms and capabilities were) I’m sure all of these issues could be overcome if I had more time. There of course are other established programs like GibbsCAM or MasterCAM. I hope to revisit this specifically in the future, pending I find a solution I like.

My simple instructions are assuming you are using Kiri:Moto and you don’t run into the issues that I faced with exporting the final gcode. So… In the menu on the left, choose your mode, in my case CNC milling. Second, choose the device, in my case it is a tinyg mill. Then, choose the tool and I have a 1/8″ end mill installed on the desktop mill which they already have some parameters for. Make sure this tool is selected in all the slicing steps you have activated on the right. So use the dropdown tabs to select the 1/8″ tool. Make sure you’ve loaded the geometry and rotated it into position. At this point, I can slice the geometry and see what the path looks like for the cutting tool. Once it completes the path generation, you can even use the slider at the bottom of the page to watch the cutting layer by layer. If you like it, then you can export the gcode for the mill. I love the little bar at the bottom of the screen where you can scroll through the roughing and finishing steps to see exactly where the tool goes.

Finally, if you can save the gcode out from whatever package you have then it’s time to setup the mill. I used the wood that I prepared yesterday and secured it to the spoiler board. I’m pretty novice at setting this up so hopefully someone will school me in the comments on the best way to prepare the spoiler board and setting up the workpiece. Basically, I moved the end mill within the approximate operating range and setup rails that I could secure the workpiece to so that it would stay in my preferred location and give me the most access to machining my features. For running the desktop mill, which is an OX CNC, it interprets gcode through a tinyg controller which is hosted by a json server. Honestly, seeing the other options I think the GBRL controller seems a little more common, but you have to walk in the shoes you own. You can move the machine with coolterm but I just jumped straight into Chilipeppr for sending gcode instructions to the tinyg controller. Again, my advice if you are getting started with this is go to Chilipeppr’s website and start pressing buttons and trying it. I was pretty impressed with the ease of use once I finally got the communications side hooked up. In the end, for the drag side of the knife pattern I started milling today as shown in the picture.

Some concluding remarks, the milling above failed today. First, the hold downs you see there were not adequate to secure the wood while the tool was traveling (thanks to Kat Steel for noticing my overconfidence). Second, my first gcode was stepping down 1mm deep each pass for the roughing cut. This stalled the machine because it was too much for the little 300W router on my system. Third, when I started the shopvac to vacuum some of the sawdust… it popped the circuit breaker on the power strip I was using and stopped the mill. So with those three interruptions, I decided to call it quits on the mill today and start it early tomorrow morning to see if I can finish something to show. Given the above discussion, I’m sure I’ll need to add some posts about setting up the machining parameters and other details but later on that.

Wood for CNC

One thing that I haven’t quite figured out is wood for the desktop cnc machine. Ideally for my purpose of quickly getting ready for building patterns, I’d like to source some wood to build patterns and teach everyone in the lab how to make larger pieces of wood from smaller ones. By the way, I’d like it to be quick and cheap too, see rapid prototyping post about difficulties with this. So, there are some nice discussion about the types of wood for pattern making in the patternmaker’s manual published by AFS. However, for my purpose I just want to go with the easiest available white pine that we get in various quantities from shipping containers or cheaply from local suppliers in 2×4 dimensions. For better machining, I’d probably get 2×10 or 2×12 but I just want to illustrate what I’m trying. Also, for this post I’m going to discuss the dimensions in inches unless otherwise specified.

Given my mold lock (which I’m going to change the outer dimensions to 10×16 for the knives, thanks for pointing this out @workshopshed), I’m ready to start machining if only I had some wood. Now, probably the best thing to do is get some 2x10s but I’m going to talk about making essentially a 2×10 using 3 sections of 2×4. For that purpose, I gathered all the miscellaneous 2×4 sections around the lab and began to cut them to 16″ lengths. Of course the horizontal band saw is probably not the preferred method, but it was effective for me. I made about 6 pieces that I can take to the planer.

The objective if I want to join them together is to prepare an edge to join them together that reduces the gap as much as possible. Don’t assume the wood is square and use a planer to create the square surface to mate. I went ahead and planed the 4″ wide section of each 2×4 top and bottom to make consistent surfaces removing any roughness on the as received pieces. Then I needed a way to secure these sections together so that I could plane them all at the same time on edge. So I built a small frame using some strut channel that we had lying around from some earlier projects and a short 12″ length of 3/8″ threaded rod. Basically, I was going for a way to hold this section of 6 sections of 2×4 together so that the planing would remove the same thickness from all to then be able to join.

Next I used a joiner to cut a slot into the side of the 2×4 for a #20 biscuit. This method is quick and easy for joining these 2×4 together on edge. Couple pointers here, make sure the depth is set correctly and that your angle is set for 90 degrees on the joiner. My joiner kept slipping below 90 degrees and the biscuits wanted to be inserted on an angle.

You can see the final product (prior to glue drying) of the wood base for my CNC milling. I used 2 biscuits between each 2×4 and the final dimensions are about 9 5/8″ by 16″. After the glue dries, this should be ready for milling. The circumference around this piece will be cut one last time at an angle to build the box for the molding once the CNC milling of the part is complete. So the fact that this is not quite flush on end and not quite flat between the 2×4 sections don’t bother me too much.

Mold Lock

In the effort to develop pattern making expertise, I was disappointed in the lack of information about various types of mold and pattern locks to align the cope and drag sand molds for a flaskless sand casting process. The most common discussion I’ve seen is the flask alignment approaches which are primarily used in green sand. There are some really good backyard casting websites that overview the process. Let me start by acknowledging them, hats off to Gingery Foundry, Workshopshed, Backyardmetalcasting, kylehausladen’s instructables, Steven Chastain’s book, AFS Patternmaker’s Manual and others I’m sure to forget to include links. Let me give the caveat that I’m looking specifically for good strategies for pattern/mold alignment that I can just build into the patten for a flaskless mold.

Let me tell you things that don’t seem to work well, to start out with. First, you could depend on the edges of the mold to align the pattern. If there are no features in the either the cope or the drag then I believe this method would work well. That’s an obvious one. Second, you could add “pins” to the pattern. In fact, there are options like the Freeman Concentric Mold-Lock Buttons that are commercially available. I think a solution like this would work well if you take care to install them both in the right location and with three of them that are not symmetric so that the orientation of the mold cannot be reversed. But frankly, I really want to discuss an alternative mold lock that I’ve been playing with. Using Onshape, I created a 1/2 inch platform that the cope and drag patterns could be built from. This is simply a “picture frame” approach to aligning the cope and drag. The drag lock is simply creating a raised surface which when matched with the cope lock which creates a recessed surface. The outer edge is 10.5″ square which is roughly the limit on the desktop milling machine. One corner has been prepared with a corner knocked off to make sure the mold is oriented properly.

Drag lock in Onshape.com

Cope lock from Onshape.com

Please note that I’m probably a little premature in claiming this is my solution for this alignment issue. But the example below is the cope side of a little cast knife project I’m working on to see if we can cast some steel knives as part of an AFS student activity.

Rapid Melting Applied to High‑Pressure Die‑Casting

… is the title of Carlos Larrazabal PhD dissertation work that he successfully defended last Thursday. Of interest was the power outage that occurred immediately after his presentation. The committee continued to question him in the dark and everyone persevered until he answered all aspects of his work.

Stay tuned for his first publication that is focused on modeling the electromagnetic and thermal aspects of induction heating a cylindrical billet of aluminum. This work is the basis for designing a rapid melting system. We believe that this effort will serve as a building block for designing several heating stages toward delivery of high-quality aluminum on-demand.

Rapid Prototyping

So, I’m working on a few things to help my “rapid prototyping” situation. As I see it there are a couple reasons that it’s difficult for me to complete patterns. So, let me discuss them below…

1. Time: this is the biggest problem and why rapid is so frustrating. If you only have 15 mins then you pretty much lose that time if you can’t get to the next milestone. You pretty much need to start over.
2. Software: what ones and what do we need? It’s not particularly clear what cad package, preprocessing steps, gcode writers, analysis, etc. is necessary to actually get a chance to machine or make something.
3. Tools: my tools are sufficient but not easy. So with much horsing around and persistence, sometimes you can get the equipment going. But most of the time, you end up back looking at the software again.
4. Finished examples: even if I found time / software / tools, the lack of examples of how to make a successful pattern, machined sample, tensile bar, etc. is a real problem when we run into difficulties.

Comment below if you think I missed anything. I’m hoping that I can take a significant leap in getting going in “rapid prototyping” where it is easy. Let me explain point by point…

1. You aren’t going to get more time (sorry). So we’ve got to find a way to make the milestones shorter to reach in the time available. So try to break whatever you are making up into milestones that you can use for the next piece. That will mean that you need to make some decisions about mounting holes, etc. early but leave yourself some flexibility so that if you need to remake that piece then it will be easy to replace.
2. We are quickly leaving the age of ProE, Solidworks, etc. most of those tools will be around, but others are taking their place (Onshape, OpenSCAD, etc.). The newer tools offer the traditional basic modeling capability but also better interoperability and data management. However, the lessons learned about how to draw engineering 3d shapes will survive.
3. Over the last 2 years, I’ve built several 3d printers and a desktop milling machine, MakerFarm and Ox CNC. These have affected my expectations just like the digital camera back in 2003 when Sarah and I got married. We had “digital photos” of our wedding day that we were thrilled to get on CD at the time. The reality is that they were scanned images of the printed photos, which for the time was still pretty awesome. However, anyone looking at those photos today would be pretty disappointed to reprint them compared to the digital images from a phone camera today. 3d printing and desktop milling machines have the same effect. I’m much more dissatisfied with my ability to cut and make something once you put that in my hands compared to before I had them.
4. Youtube, Thingiverse and GrabCAD have now made finished examples easy to get and use. The problem is that even with all the examples, the ones I’m looking for don’t exist.

My conclusion is that I think the time is right to see progress to being able to propose and think of a thing and then make it. However, until I reach a critical mass of 3d models and worked examples for myself and my lab then I’m still going to be whining about this rather than just making stuff.