Christopher Kalitin Blog

Claude For Solid Mech Analysis of a Mojave Sphinx Test Stand

Nov 19, 2026 - 3165 words
Christopher Kalitin Blog (https://ckalitin.github.io/posts/sphinx-test-stand-fbd/)

UBC Rocket

Image

Recently I joined the UBC Rocket student design team. We’ve previously won first place at Spaceport America, static fired multiple liquid rocket engines, and have alumni working at all the space companies you know.

Recently, we’ve started a new program to make a Half Cat Mojave Sphinx liquid rocket and static fire it in the coming months. As part of this program, we assembled a test stand (shown above) in about a week from Aluminum Extrusion we had in the shop. By vibes and pushing hard on the test stand it looked like this test stand would work well, but I had some free time on a flight to the sphere recently so I tested Claude’s ability to do Solid Mechanics.

Overview

High level model overview:

All code for this model is available on my GitHub (pending moving it to a UBC Rocket repo).

Claude doing the math.

Triangle Bracket Free Body Diagram

Image
Representative 8020 triangle bracket
Image
8020 Triangle bracket with an anti-rotation key that slots into the 8020 channel - McMaster
Image
Whiteboard proof of work, I didn’t just blindly listen to Claude here!
Image
Claude’s diagram of the reaction forces on one bracket joint.

Each triangle bracket has two screw joints. Both of these joints can resist forces and moments in x,y,z.

Forces:

Moments:

Because each triangle bracket has two joints, there are pairs of forces for each bracket.

For example, referencing Claude’s diagram above, positive Z force on member A is resisted by the screw in tension whereas the same force on the member B joint is resisted by friction against the bracket and the aluminum extrusion.

Maximum Allowable Forces & Moments

Assumptions:

Forces:

Moments:

Image
Claude’s max allowable moment diagrams.

Joining Plate Free Body Diagram

Image
Claude’s Diagram, given my Onshape bracket CAD.

The joining plates are a special case of the brackets where the same joint is used 5 times on the same plane, instead of twice on planes that are 90 degrees from each other. Note that the joints are split into two groups, A and B, depending on the member they attach to.

Because there are multiple joints per member, we need to solve the indeterminate forces to find how much of each force is distributed along each joint.

Image
Claude’s diagram of how load splits across a bolt group.

Each bolt group is in line, so the split has a closed form. Force is shared equally. Moment is shared in proportion to each bolt’s signed distance from the group centroid (illustrated in the first diagram of this section), acting perpendicular to the line. The centre bolt of a 3-bolt group carries no moment.

Maximum Allowable Forces & Moments

Assumptions:

Group A (3 bolts in member A, at r = −25.4, 0, +25.4 mm from the group centroid):

Group B (2 bolts in member B, at r = ±12.7 mm from the group centroid):

Claude did the math above.

Model Parameters

Model variations:

By using the worst-case values from the pin joints, rigid joints, and non-continuous member cases we can find the most conservative estimate of safety factor.

Member failure will be modeled using:

Image
CAD export with every angle bracket and plate bracket.

This diagram shows an export of the CAD along with every angle bracket and plate bracket. This model is available here.

Model Validation

In the spirit of the great Graham Hendra, I backtested this model using his Burrard St. Bridge worked example. He models the bridge as a truss with pin joints and examines the members in buckling.

Image
Graham Hendra’s MECH 360 Burrard Bridge worked example.
Image
Method of Joints member loads from the worked example.
Image
Yield and Euler buckling safety factors from the worked example.

Burrard Street Bridge Validation Results

All values in my model are equal to the provided 3 sig fig values in Graham Hendra’s model. The model passed this back test / validation.

Image
My model’s member forces for the Burrard half truss.

Member Forces

memberN (MN)N/Pnotes N/Pmatch
HJ-10.875-7.250-7.250yes
FH-10.458-6.972-6.972yes
DF-9.790-6.527-6.527yes
AB-8.644-5.763-5.763yes
BD-8.553-5.702-5.702yes
DE-1.292-0.862-0.862yes
FG-0.330-0.220-0.220yes
JK+0.000+0.000+0.000yes
HI+0.291+0.194+0.194yes
HK+0.884+0.590+0.590yes
FI+1.453+0.969+0.969yes
BC+1.500+1.000+1.000yes
DG+2.309+1.539+1.539yes
BE+3.949+2.633+2.633yes
AC+5.400+3.600+3.600yes
CE+5.400+3.600+3.600yes
EG+8.192+5.462+5.462yes
GI+9.600+6.400+6.400yes
IK+10.406+6.938+6.938yes

Member Buckling Safety Factors

memberSF yieldnotesP_crit MNnotesSF bucklingnotes
BE2.712.71
HJ3.033.03232.92232.9221.4221.42
FH3.153.15230.62230.6222.0522.05
IK3.17
BC3.213.21
DF3.373.37223.96223.9622.8822.88
GI3.443.44
DG3.663.66
DE3.723.725.845.844.524.52
AB3.823.88142.03142.0316.4316.69
BD3.863.86213.69213.6924.9824.98
EG4.034.03
FI5.825.82
AC6.116.11
CE6.116.11
HK9.489.47
FG13.0013.002.632.637.987.98
HI16.5516.53
JK—

Model Diagrams

These diagrams of the 2D and 3D models define all nodes and members.

2D Model

Image

3D Model

Image

2D Model

Finally, we can solve the system that corresponds with our test stand. We will apply a 250 lbf (1112 N) upward force on the member that corresponds with the rocket, then calculate the force and moment on each joint, and finally determine each safety factor.

Below, we see that the lowest safety factor is 0.08, and the system fails. The limiting factor is the Z-axis moment that must be resisted by friction on each joint.

A safety factor of 0.08 suggests that a force of just 20 lbf would cause the stand to fail (safety factor = 1). We’ve applied more force than this ourselves, so before we continue we must define what failure is occurring and discuss if the model is valid.

Image

Joint failure in Z-axis moment means that the joint starts rotating as shown in the left diagram above. This failure means that friction force no longer resists the moment and the members move until they settle into a new equilibrium.

In 2D, this means that we get rotation of the joint. Given the Free Body Diagrams we constructed of the brackets, this 2D behaviour clearly doesn’t fully capture what we would see in 3D. So, we must model our test stand in 3 dimensions.

Image
2D model safety factor by member and joint.
Image
2D model lowest safety factors.
Image
2D model deflected shape (deformation exaggerated 11x).

This extreme deformation is also something that we should look into. In the next section we’ll see that it is not present when we model this problem in 3D.

This 2D modelling section has been a lesson in being careful what simplifying assumptions you make.

3D Model

Our 3D model still has a minimum safety factor of 0.08, but on an entirely different set of joints.

In the charts below, notice that joints on the ground (N5, N7, N6, etc.) have the lowest safety factors. This is because the bottom of the test stand does not consist of continuous members, instead it is made of ~10 sections of aluminum extrusion because that is what we had on hand.

Every joint between these disparate sections of aluminum extrusion carries a high moment (because they are offset ~1 inch from each member, so they’re not directly in the force path). This leads to the high Z-axis moments that cause these low safety factors.

So far it’s not looking good for our test stand design at all!

Image
3D model safety factor by member and joint.
Image
3D model lowest safety factors.
Image
3D model peak combined stress.
Image
3D model ground reaction on the base.

To get the distribution of ground reaction forces this model used an array of high-stiffness springs on the surface to find where the most tension and compression lies. More implementation details are available in the GitHub repo.

Anti-Rotation Key 2D & 3D Simulations

Image
8020 Triangle bracket with an anti-rotation key that slots into the 8020 channel - McMaster

In the previous simulations we found that the limiting safety factor was Z-axis moment on each joint. A potential solution to this problem is using triangle brackets that have anti-rotation keys. These slot into the 8020 channel and allow us to resist moment with normal force on the inside of the channel instead of purely friction.

For reference, we previously had a maximum moment of 4.2 Nm and now we’re up to 36 Nm.

Image
3D model lowest safety factors with anti-rotation key brackets.

In the chart above we can see that Z-axis moment is still our limiting factor. However, other failure modes like slip are relatively more important as they are now on the same order of magnitude of safety factor.

Conclusion

The current design of our 8020 aluminum extrusion test stand will fail during a static fire and cannot be used.

This model doesn’t perfectly reflect reality, and in some ways it is conservative and in others it underestimates. For example, in some cases the Z-axis moment would be handled by members coming into contact with each other and transferring the moment through normal forces. An example of underestimation is that the model didn’t include any off-axis thrust forces, it assumed the thrust force is purely vertical.

Potential avenues for redesign of the test stand include:

Download PDF