Curved Plywood Shelves

Inspiration

This project has been bouncing around in my mind ever since the Innovation Center opened at Dawson in 2020. When it came time to order office furniture for the building, I decided that I would not order any shelving with the intention of someday creating my own shelving design using the equipment in the Makerspace. After many iterations of ideation, I landing on this "bent" plywood design which originally came out of a sketched created by my friend Jeremy....

Ideation -> Concept -> Adapt -> Design -> Iteration -> Creation

2/28/25 - The photos to the left are from my sketchbook. Several of these sketches we early attempts at coming up with a pleasing concept for the shelves. For almost 5 years, I thought about what these shelves might look like, but never really landed on an idea that I liked. On NYE 2024, I was talking with Jeremy about this project and just as he was about to leave our houses for the evening he quickly drew this pair of wavy lines in my book...

When referring back to the book a few days later, I thought about how to adapt the top line to something that might work as a shelf and have horizontal surfaces that could hold books.

Note - I have had several cardboard boxes of books, objects, and artwork/creations that I've been wanting display stashed away in boxes around my office since I moved in.

Looking back at some of these early sketches, now that I've created the front shelf, I realize that some of these ideas might actually be possible to create. I am particularly interested in what it might look like to create wall mount flower similar to the sketch I created towards the beginning of this journey.

It's fun to look back at these sketches now that I've made the front shelves and see how creating these simple sketches evolved into their final created form.

CAD - Fusion 360

3/1/2-025

My intention for the CAD for this project was to design as robust of a model as possible so that it would be as easy as possible to make adjustments with out breaking features in the model. This began with thinking about how the shelves would fit in the physical space. I modeled the parts of the two walls in the selected corner of my office so that I could visualize how it would fit with existing elements (my desk, window, mounted guitar, ceiling height, etc).

Once the walls were modeled, the first major and key step was to create a well defined sketch for the first shelf on the front wall. I created some user defined parameters that would be easily adjustable and included those parameters (rather than setting dimensions) as much as possible when creating the sketches. The parameters included the thickness of the stock material I planned to use, a standard radius for all of the major curves in the design, and the thickness of the shelf.

As the project progressed, I went back and added or modified parameters to allow the design to evolve to solve issues that I discovered while modeling. At first the plan was for the whole shelf to be one thickness across all of the layers of the plywood to create a consisten thickness for the shelf. I eventually realized I could save material and reduce the overall weight of the finished product by making the front and back layer thicker than all of the layers sandwiched in between. By separating the original parameter for shelf thickness into frontBack and middlePieces, I could experiment with various configuration to see what looked best and also fit the envelope of one sheet of 8' x 4' plywood (.92" thick).

I eventually realized that slicing up the middle layers into smaller pieces would allow me to fit more parts into the 8' x 4' sheet. This led to many revisions as I explored ways to slice up each layer and ultimately assemble the parts that make up the composite shelf. I ended up with 38 individual pieces to create the front shelf.

Fusion has a very powerful tool called Arrange that allows a user to select an envelope to contain many components. This makes it possible to model the shelf as it will be assembled, but break out the parts as they will be cut on the CNC. The Arrange function accounts for the size of the envelope, the desired components to fit inside the envelope, and spacing between components and the edge of the envelope. The Arrange function automatically recalculates the placement of selected components when parameters are changed, which is very handy! This means I can change the thickness of the inner layers and see how that impacts placing the components in the stock plywood.

I was able to play with my parameters until I could efficiently fit everything for the front shelf all in a single sheet of plywood. I could have sliced the pieces into smaller parts to be even more efficent, but the downside would be the increased difficulty of assemblying the final shelf. I ended up coming up with a version of the shelf that removed several pieces from the design to create a pattern that reminded me of a park bench with gaps between each section. I chose to create a repeating pattern of missing pieces to create the final design which looked appealing to me, made it possible to fit the parts in the 8' x 4' envelope, and further reduced the overall final weight of the shelf.

Considering the installation - stud finding, model adjustments and sketch modification/

Manufacturing

3/1/2025 - Once the CNC routing was completed, the next step was cutting out each of the individual 38 pieces using a jigsaw. The pieces are held to the main plywood stock using 1/4" tabs. The jigsaw cuts through the tabs but leaves a section of each tab along the perimeter of the pieces. These are removed using a flush-trim bit on a router table. Each piece must be cleaned up with the flush-trim bit to remove all of the tabs and then the pices are individually sanded along the edges and on the faces. An occilating sander helped speed up the process of sanding the curved pieces.

Planning for mounting

As I was in the process of modeling the shelves, I was fortunate to think ahead to how the shelves would be mounted. Thinking ahead gave me pause to slow the design until I made some choices about how they would hang on the wall. I found some metal French cleat hardware online that was very low profile but strong enough to hold over 100 lbs. The two pieces of the French cleat only take up about 1.6", so I designed the shelving to be slightly wider than that so the hardware could be completely hidden behind the shelving. Using parameters in Fusion 360 made it easy to change the overall thickeness of the front and back pieces of the shelf and I settled on 1.75" for this parameter. The sketches and 3D features of the model were intentionally modeled to be parametrically driven and robust enough to handle a wide variety of ranges for essential aspects of the shelves.

I used both an electric stud finder and a strong magnet to determine the location of the metal studs behind the drywall in the corner of my office. I draw them in pencil on the wall and eventually modeled them as as sketch in the Fusion 360 project. This allowed me to precisely determine the locations for each of the French cleats and led me to make some small changes to the design to make sure the cleats would go on the shelf in key places that aligned to the studs.

Several of the metal French cleat pieces needed to be shaped to fit the curve of the shelf. I was able to remove small corners of the hardware in a way that did not compromise the strength of the cleats and screws, but still allowed the hardware to be hidden. I also had to modify the hardware by drilling countersunk holes to accomodate the heads of the larger metal tapping screws I purchased for the project.

One final challenge in using the French cleats was developing a plan for how to precisely align the cleat receiver hardware on the wall so it matched their respective cleat hardware that was installed on the back of the shelf. I decided to print a mounting template on our large format printer and hung it on the wall to determine the exact location to drill holes for each cleat receiver. Then I was able to remove the mounting template print out, drill the rest of the holes, and mount them with the metal tapping screws or with drywall anchors for spots where screws didn't align to studs.

The back piece was test fit to the cleat receivers to confirm everything was aligned prior to beginning assembly of the rest of the shelf.

Nar also suggested wrapping plwood scraps that matched the thickness of the wood layers using wax freezer paper to help avoid glue sticking to the spacers. This worked really well. Ultimately the biggest mistake was getting glue on the inside face of the back piece of the shelf which is visible from the front of the shelf. This is something I will be more careful about moving forward on the other shelf and was cleaned up with extra sanding.

Once the first five layers were glued and clamped, we added four more layers the following day followed by the final four layers the day after that. The picture at the right is all thirteen layers clamped for the final shelf. After the glue set, I spend almost two weeks sanding the shelfs to remove glue squeeze out and make them as smooth as possible. Some areas of the curved sections were quite difficult to sand and I am considering an alternative method of assembly for the second shelf that would use much less glue. We have been considering using threaded rod instead of metal dowels to hold the shelving together. This would eliminate the use of most if not all of the glue. The downside would be that the threaded rod will need to be tightened with metal nuts and these will be visible unless intentionally hidden. I am considering using threaded rod for all of the layers except the front facade. Then I could create a pocket in the back side of the front piece that hides each nut on the second layer from the front.

Shelf Assembly

Putting the shelf together was a multi-day process. In retrospect, I wish I would have been more cautious and methodical in the glue work as the first stage of assembly ended up a bit messy and led to more clean up later. After cutting and installing the 1/4" metal dowels in the back piece, we glued the first five layers of the shelf. My colleague, Narciso, suggested that we add spacers in each of the gaps where there would be no wood, so that we could confidently clamp the layers together without worrying about warping or bending unsupported layers.

Front shelf installed and in use in my office office.

There were several major challenges in assembling the first set of shelves that I am hoping to solve in this second set. One of the biggest issues was the amount of glue that got on the shelves which required lots of clean-up and sanding after the shelves were put glue and clamped together. To solve this issue, I am trying to use 1/4"-20 all-thread rod (instead of plain steel rod) with square nuts on each end of the rod to hold all of the layers together. My goal is to only use glue to attach the front facade piece to all of the other layers once they are assembled bolted together.

The plan is to put the rod through the back piece that will be mounted to the wall with the French Cleats and use pairs of square nuts to tighten the 12 layers of wood together. The piece that is at the front of the stack (just before the front facade) will have two nuts sticking out in front of that layer. Then the front facade will slide over all of the square bolts by aligning with a milled square pocket fot each set of nuts.

Phase Two: Left Shelves

March 30, 2025

For the second set of shelves designed to hang in the corner of my office, I wanted to build off the aesthetic of the original shelves and create a complementary design to add more wall storage for books, artwork, projects, etc. The mockup to the left is the design I settled on. It includes the same type of layout with some of the pieces being removed to create a more airy feel. The missing pieces lower the weight and material use for the project with the benefit of making it possible to hang objects underneath some sections using cord or twine.

Top view of the threaded rod as it passes through the back 12 layers of plywood and is attached with the square nuts on both ends. Note: there are three large sections of the shelf on the far right edge (near the corner of the office) that looked like they could use additional strength. I have design small pockets in three groups that each will have a small section of steel rod inside. I may use a small amount of glue to help hold these sections together as there is no way to use threaded rod and nuts here without them being visable. I am hoping that by using pairs of square nuts, I can keep them from gradually loosening and losing some of their clamping pressure. Again, my goal is to use little to no glue for the majority of this assembly. If there are noticeable gaps between sections of plywood, I may need to add some glue.

Top-down view of the corner with both shelves modeled.

Finishing pass simulation. 0.05" of stock to leave is being removed all in one full-depth pass.

I made an additional adjustment to make the parts mill more cleanly. I added a finishing pass after milling the 2D contour around all of the parts. I turned stock to leave on with 0.05" inches of radial material left for the finishing pass to clean up. This is a technique that we experimented with in our guitar making workshop over Winterim this year and it really helped reduce the tool marks that are left on the edge of all of the parts. By duplicating the original 2D contour with tabs enabled and toggling off stock to leave and multiple depths, we should be able to get a much cleaner contour cut around all of the parts which should reduce sanding and clean up.

Another improvement that I wanted to make was using Fusion 360 and the CNC to label each of the parts as they are milled to make it easier to identify each of the pieces and simplify assembly. This ended up not being as straightforward as I hoped. I eventually was able to create a sketch with text elements with the name of each piece and get the sizing of the font to fit within each section of milled wood. I experimented with the engrave feature in CAM but it didn't work as desired. I eventually landed on using the trace feature instead. This required offsetting the sketch to the depth that I wanted to mill the text labels. I only need a very small mark with the label for each part, so I offset that sketch to a depth of -0.05" below the surface of the stock. I think I will be able to mill the labels with a 1/8" end mill.

Simulation of the completed CAM in Fusion 360. Note the small 1/8" holes in the corners of the square pockets. This should make the square nuts fit smoothly.

Page updated

Report abuse