ME 424 Capstone · Senior Year · Spring 2026
Ballistic Gel Slicer
A coordinate-driven slicing machine for the U.S. Army DEVCOM Armaments Center. Live-fire ballistic testing leaves blocks of gel full of fragments that need to be located and extracted; the current workflow is a technician with a knife. We built the machine that automates the slicing half of that.
Team of four (Alexandra Bergman, Pierre Gathy, Fernando Rodriguez, Giovanni Smith). Mechanical design was a shared effort. I owned the electronics and firmware — wiring, motor drivers, Hall-feedback synchronization, and the Arduino command interface that runs the cut sequence.
What it is
The Army's Test & Evaluation Division shoots projectiles into transparent ballistic gelatin to study how they fragment. Each block is then dissected by hand — a technician slices it at intervals, finds each fragment, measures and extracts it. It's slow, inconsistent between operators, and the bottleneck on a 500-block test campaign.
Our machine takes that knife-work off the technician. The operator types in the cut coordinates and the system positions the block, drops a guillotine blade through it, and retracts — repeating for every cut on the list. Same workflow as before, just hands-off and consistent.
How it works
The structure is an 80/20 aluminum frame with a stainless-steel bed. Two NEMA 17 stepper motors (StepperOnline 17HS26, 79 N·cm holding) drive parallel Tr8×8 lead screws in lockstep, pushing an aluminum plate that translates the gel block along the x-axis. With 16 microstepping on DM542T drivers and an 8 mm lead, that's 400 steps/mm — about 0.0025 mm of theoretical resolution per step.
The cut is a guillotine blade driven by two Progressive Automations PA-09 linear actuators (12-inch stroke, 330 lb rated, integrated Hall-effect feedback at 660 pulses/inch). Two actuators rather than one centred actuator so the cut force is even at both ends of the blade and it stays parallel through the gel.
Control is an Arduino Uno R3. The stepper drivers share a STEP/DIR pair (the lockstep coupling for the lead screws); each linear actuator gets its own BTS7960 H-bridge so the firmware can throttle one independently when the Hall-feedback synchronization error between the two exceeds tolerance (12 pulses, ~0.018 in). That keeps the blade level even when one side fights more material than the other. A 24 V / 20 A switching supply feeds both the steppers and the H-bridges off a common rail.
The operator interface is plain serial at 115200 baud — no app, no screen. Commands are the obvious
ones: ZERO, LOAD x1,x2,…, RUN, STOP,
STATUS, plus single-axis test commands for bring-up. The full pin map and the firmware
file are in Appendix G of the report.
On the engineering-analysis side, we sized the motors and lead screws against measured friction (μ ≈ 2.24 between gel and stainless) and measured cutting force (~21 lbf for a guillotine blade), then ran FEA on the push plate, blade mounts, and base plate. Early PLA versions of the push plate and blade mounts failed in mockup testing as predicted; the final build switched both to aluminum.
Where it's at
Working prototype, delivered to the client at the end of the semester. End-to-end cycle time of ~7 minutes per block beat the >50% labor-reduction target. The dual-actuator Hall-synchronization held the blade parallel through every test cut on cured gel.
Outstanding before it can ship to a live-fire range: a full blade guard, an electronics enclosure (the drivers and Arduino are currently exposed on the bench), routed cable management along the 80/20 channels, and through-bolting the actuators rather than relying on friction at the PLA mounts. Next step on the science side is automatic fragment localization so the machine can generate its own cut list rather than taking one from the operator.