Work Experience

An automated tire inspection machine used a dedicated arm to eject scanned tires and separate mechanisms to catch incoming ones. The goal was to do both jobs with one arm: make the eject arm extend on pneumatic cylinders so that, in its raised position, it can also meet an incoming tire and cushion it by extending along the direction of travel. I helped design and build the first prototype around linear slides, twin 3/4" pneumatic cylinders, machined aluminum components and custom sheet metal covers, then assembled it and installed it on the machine. It worked on the first test, and the project became mine to develop from there. What made it interesting was the sequencing. Arm angle is driven by an electric cylinder while extension and retraction run on solenoid-valve pneumatics, so the catch, the cushioned descent and the return to park all have to be timed against a tire that is already rolling. Getting the release timed so the tire meets the arm at the earliest safe moment was where most of the cycle time came from.

Validating the arm's performance needed data that did not exist yet, so I built the test setup before I built confidence in the mechanism. The rig runs on an Arduino and integrates LiDAR and ultrasonic sensors, a load cell, a pressure transducer and a custom linear position sensor I made by driving a rotary potentiometer through a slider-crank linkage. I made up every cable and connector as sensors were added. On the software side I wrote a Python desktop application that captures each run over serial, plots it live, computes the metrics and generates test reports automatically. It replaced a manual workflow of reading a terminal and pasting numbers into a spreadsheet. The most useful result was a negative one. I had expected closed-loop cushion control to beat a fixed valve opening, and the data showed it did not. We kept the simpler design.

A single-degree-of-freedom arm sweeps a scanning head into and out of a tire, and its trajectory is set entirely by a cam the follower rides on. The original profile caused aggressive jerk, unwanted vibration, and at higher speeds the follower would leave the cam altogether — which capped how fast the arm could be driven. I redesigned it as an optimized polynomial motion profile that minimizes both acceleration and jerk, modelled it in SolidWorks, and proved it first by machining the existing cam to the new shape. The improvement was immediate: the vibration and separation disappeared and the arm could be run as fast as the actuator allowed. The profile has since been adopted into the product. The harder half was manufacturability. The cam follows a cylindrical path, so the surface is three-dimensional. I reworked the geometry until it could be machined from a small piece of aluminum flat bar, and printed a 3D version along the way as a functional check.

I was given a product idea to look into a handheld laser caliper device and asked to find out how to make it work as a viable product. I ended up building the whole first prototype with a viable bill of materials.

The core problem was interfacing a linear CCD with an Arduino, which meant generating multiple precise timing signals and proving them on an oscilloscope. I had close to no microcontroller timing experience going in, and spent a solid stretch of the term learning it properly.

I also applied some optics using a lens to collimate the laser light before reaching the sensor.

Once the system worked on a breadboard I compacted everything onto a hand-soldered perfboard assembly and designed a 3D-printed enclosure around it with battery power, an e-paper display and a custom tactile trigger.

An industrial scanner was being installed in an environment dirty enough to put its optical glass at risk, and a shield had to exist in time for the unit to ship. I started with cardboard mock-ups to understand the airflow, then designed the real part as sheet metal in SolidWorks, sent it out for fabrication, and fitted it on return. Working to a fixed ship date meant the design had to be right the first time rather than iterated, so the cheap cardboard stage earned its keep.