Links

At Subaru, performance evaluations are structured around a bell curve distribution, where "Meets Expectations" represents the median rating and is generally considered a positive assessment. That said, my evaluation reflects both the impact of my contributions over the past year and my professional capabilities. A complete, unedited version of the review is available upon request. This evaluation was prepared by my direct supervisor and his manager, reflecting a joint assessment of my performance.

Shown in the photo above is a CAD model of a front door exterior panel with a real-world photo shrink-wrapped onto it. In our manufacturing process, sealer is applied to this panel in locations that align with other parts of the door. At this stage, it is difficult to determine whether the sealer has been applied accurately due to the lack of distinguishing features on the panel. By overlaying the real-world photo onto the CAD model and comparing it to the drawing, the accuracy of the application can be predicted, thereby reducing the number of corrections and countermeasure cycles needed for precise application.

Digital simulation helps to ensure the effectiveness of designs and recude downtime via visualized communication and digital twinning. The moving part shown is part of a hood that is placed into the stand by a 6 DOF robot. The stand must be capible of hold three different varities of hoods, and able to differentiate which model is loaded via its attached sensors. By loading the CAD model of each part into the assembly, the operation of the sensors can be ensured. Furthermore, by planning the motion of the part, it can be guaranteed that the handling robot is able to enter and exit the tight space.

The gold-colored part shown above is manufactured in-house and is loaded onto the light-grey rack by a robot. The part's orientation requires the operator in the next process to rotate it 180 degrees to load it correctly onto the jig. This extra handling step can increase the likelihood of the part being damaged or the operator getting cut by a sharp edge. The newly designed transfer rack (dark grey) is used to reverse the part's orientation before it reaches the operator. It is currently being prototyped and is expected to be placed into service over the next few weeks. There is a single part loaded on the rack in the animation, however the fully loaded rack must support a weight in excess of 1000 lbs. FEA analysis was conducted on the rack, with an estimated FS of ~8. (It was determined that the automated loading process could not achieve the desired part orientation during initial loading.)

Every week I lead a meeting with our new model development (pilot) team. This meeting includes members of production, quality, and other support areas, with the total number of attendees usually being around 15~20. While being sure to not drag in unnecessary audience members, I believe inclusiveness and transparency have helped me build trust with coworkers. I think that I am able to have a fairly large audience every week because my messages are well organized, I keep the meeting focused, and those who attend are able to speak their opinions, have their voices reflected with action, and be well informed of current activities. Click the image above to see an example of the sides I use to communicate with my team!

As we begin integrating new equipment, starting dry runs, and building new model parts, design flaws and limitations which were not identifed in the design phase start to stand out. For large issues, we have to work with the jig maker and often several other parties, but for the small to medium issues the design work falls to me. Pictured on the left, is a nylon block that the operator set the door outer panel on, we are currently working ona new design to help the making placement easier, with less margin for error. On the right is a locator for an internal panel in the door assembly. The orignal design allows for the left hand side part to be placed on the right hand side jig upside down (and vice versa). This design change, which we catagorize as pokayoke, is aimed at preventing the ability to misplace the part in the described manner.

Lately, I've been working with Mitsubishi PLCs and HMIs to control our production lines. I'm confident in integrating hardware, troubleshooting, and making modifications. Additionally, I have experience with Omron PLCs and some familiarity with Allen-Bradley systems.

Having had done cycle time evaluations before, I admittedly (at first) underestimated the complexity of measuring cycle times. With production targets, available working time, and estimated downtime data avaiable, setting a target for the line speeds was fairly straight forward. The complexity seems to have arrisen from lack of an established method for measurement, and from the vastness of stacked, dependant processes. To overcome the challange, I evaluated what worked well and what did not work well of past studies, made adjustments, and finally started taking measurements from videos, recorded timestamps, and integrated notes justifying start and stop times used for evaluation. Moving forward, I have established a template which is already being implemented in a different area as we speak.

This photo is one example of a poster I used for an obeya meeting. Each section of the poster (except the bottom center) includes data I track and update weekly. This data is usually used for informing our weekly activites and project process, but it also allows for a quick assembly of the KPIs for our area (and in this case, presentation to management).

The graph above is the analysis of the results from my co-authored CHI paper. I have experience designing trialS to evaluate User Experience (UX), using standard testing methods including System Usability Scale (SUS), User Experience Questionnaire (UEQ), Virtual Reality Sickness Questionnaire (VRSQ), and NASA Task Load Index (TLX).

As part of my research projects at Purdue, I learned how to use Unity to build enviornments in which to test haptic devices in VR scenes. The left half of the picture above shows a free scene downloaded from the Unity Asset Store, which was then developed into a 'treasure hunting' game/research project. When paired with the haptic device shown in the gallery below, users would receive cues when nearing a target. For this project, I wrote C# scripts to manipulate objects in the scene, communicate with wireless haptic devices via TCP/Wi-Fi protocols, and measure user performance. I also used Blender to create and modify some objects to hide targets within the scene.

Robot U is a 2 week Subaru internal training program to increase familiarity with our two most common types of 6 DOF robot brands, Yaskawa Motoman and Nachi. Throughtout the program, we learned how to manipulate the robots in difference coordinate systems, create, save, and modify programs, and best practices for robot teaching.

Wearable haptics when paired with VR provide a physical sensation to the user to enhance their sensory experience and create a more immersive interaction with digital environment. I created this wearable device which was used in a study to examine how users would respond to different cues in an object search scenario. Four of these devices were worn (2 upper body, 2 lower, left and right sides) to guide users to hidden objects in a treasure hunting game. The underside of the device holds an ERM motor to create a vibration, while inside is housed a power switch, a battery, and an ESP32 Beetle microcontroller for controlling the device and communicating with the VR device via Wi-Fi UDP communication.

The final project from this class asked us to incorporate several different topics from class into a project of our choosing. I made a prototype for an automatic plant watering system. It used a capacitive sensor to measure the moisture of the soil, and when below a certain level trigger a watering sequence. When watering the servo opens to allow water to pass, and also sends a message to a server microcontroller via a nrf24l01 radio card. The server communicates with python via telemetrix to keep a watering log on the computer. There are no mechanical fasteners or glue used; It is held together solely by geometric tolerancing.

As part of this class project, myself and a teammate built the robot above. It worked to navigate a 7’ x 7’ area, detect the color on the ground at 3 pre-programed positions, light a LED strip with the corresponding color (I2C serial communication), while displaying the robots position coordinates via two 7-segment displays.

In the class I TA for, the midterm guided project is for the student to build a remote control car with an ESP32 micro controller. I built one that featured an interchangeable gear system for a race and battle mode. My race time has not been beaten so far by 64 teams of students (the battle record is a different story).

At the end of the semester students present their action toys which incorporate aspects of 3D printing, laser cutting, and design. Each toy is required to meet a certain level of complexity which is agreed upon by the student teams and teaching staff. Toys that win awards for engineering, marketability, design, and innovation are display in the engineering building atrium.

As part of a course objective, using a software called Flexium 3D I wrote G code by hand and simulated the tooling path. The part was then cut in the school machine shop.

While working at NSPA, there were around 30 cutting machines under my responsibility, but this one was particularly challenging. It was a source of operator frustration and customer complaints. To increase the reliability of the machine I rewrote the entire PLC code, redesigned the HMI (shown next), and worked with a 3rd party contractor to redesign and build the transfer mechanism. There was much testing, and over time I built a strong relationship with the machine operators.

HMIs provide an insight to what the machine is thinking and doing. Designed properly, it can help the operator to work safely and efficiently. It can also help engineers and maintenance workers troubleshoot and retrieve information without connecting to the PLC. I designed this home screen to most clearly show the operating status of the machine, the faults and conditions that enable/disable the machine to run, and easy navigation. Clickable links are underlined to enable fast troubleshooting.

A hydraulic cylinder directly related to the feed speed of the initial rough cut of the long stock steel tubes was regulated by an unregulated flow valve. Due to this oversight, precisely controlling the feed rate of the cutter was not possible and varied from machine to machine. Using a DAQ system, I measured the feed rates during the different stages of cutting so cutting conditions could more accurately replicated. After measuring the other major components of the machines, I developed a formula that was around 25% more accurate to estimate job times than the previous method.

One major project I worked on was improving the cutting conditions of 304 Stainless Steel. Due to the process used by the cutting machines, there is a very limited availability of stock insert holders that fit into the machine securely. As a result, they had to be design and fabricated by scratch. The image above was used in a quarterly progress report with the engineering manager and company president highlighting the critical design change.

Building off the last image, for a high volume SUS304 parts, all three of the cutting insert holders were redesigned to feature a positive cutting angle. All parts required tight tolerance control. They were entirely designed by me and fabricated by a third party.

Particularly for stainless steel but also for carbon steel, cutting insert trials were conducted on a regular basis looking for the most cost-effective parts as they played a major consumable cost across my departments. Examining the failure method on inserts could help to determine if the insert needed to be harder or tougher.