Hi there! 👋 I love FEA, CAD, Product Design, and Project Management. Let me show you what I enjoyed working on lately!
The first project to cover is the Effective Armored Guard Layer Enclosure (E.A.G.L.E.). The protective enclosure project is related to HYMH lab's Instron Fatigue Testing Machine. From the factory, it does not come with a protective enclosure, but due to safety concerns, the decision was made to equip it with one to protect users from shattering material.
There is already a protective enclosure in the lab. However, It needs to suit lab needs better. It is brittle, unnecessarily complicated to use, provides insufficient protection, looks unprofessional, and is not secured to the machine.
One of the only manufacturers that can produce protective enclosures for the machine is the machine manufacturer. They provide a safety cover with unique features such as an E-Stop locking mechanism; it provides an uncluttered enclosure, but the control unit is located outside of the machine boundaries, which, due to the machine's location, is a disadvantage. Due to the high cost of $17646 and lead time of 16 weeks, the decision to design a protective enclosure ourselves was made.
I overtook the project and led it from development to manufacturing. The first step of successful product development is to understand the problem you are trying to solve with it. I went to the lab and got my own user experience of the old protective enclosures. I interviewed every lab worker on their experience with the old design and what they wished to have in a new design without thinking about what could be realistically implemented. After I collected all the feedback needed for research, I started to figure out what restrictions I had. My main restriction was the dimensions of the machine. I created an accurate CAD model on which my design was based. After analyzing the feedback, I outlined the main targets that must be met. The main requirement was to have protection for the full range of machine operations. The first thought that came to my mind was to make a protective enclosure with a height equal to the maximum fatigue testing system height. After critically thinking about my first decision, I determined a disconnect between the target and the implementation. Instead of putting everything behind the safety panel, my target is to have the specimen behind the protective enclosure, and that's only what matters. I accurately calculated the perfect height of the design that would be high enough to cover the specimen but also low enough not to be destroyed by the machine in the lowest position. Using these research conclusions, I created a new design for Siemens NX. The main focus was to accomplish all the goals given.
I wanted to add an accent to the user experience. The old safety cover was unnecessarily difficult to use since the user needed to lift off and put on the safety cover on the machine each time, and this was not only time-consuming but also challenging due to its weight. My solution is to use one single door for easy everyday use, and it also has lift-off hinges to allow removal in case an oversized item needs to be tested. One unique thing that catches attention immediately is the difference in the color of the handles. This detail highlights that the user should open the door horizontally rather than lifting off from its hinges with each use.
After I finished the design, it came time for the manufacturing development, such as figuring out what parts and how I would manufacture and assemble them. After thorough research, the 8020 Aluminum profiles were chosen as a frame due to their low cost, high customization, and simple assembly. The next thing to decide on is the panels' material for it. The previous design had an acrylic plastic panel, which is a popular choice for panels where transparency is essential. However, I decided to research all materials that suit this project's needs and determined that polycarbonate is a superior choice because of its protective properties. For example, polycarbonate panels have 40 times higher impact strength than acrylic panels.
After assembling all the parts together in Siemens NX, I created the bill of material, received a quote from the contractor, and successfully installed the E.A.G.L.E. on the machine.
The next big project is Automatic Weld Efficiently Saving Operational Minutes Effectively (A.W.E.S.O.M.E.), aka Weld Automation Project. The Global Modeling and Simulation team at HYMH has a significant backlog of requests for FEA and a finite number of resources to perform analysis. I found that modeling welds for durability analysis is critical but time-consuming. The new ANSYS update delivered a new weld function, and my project was to investigate it and determine its implementation in the workflow.
Here is the procedure to execute Seamweld Analysis: you get raw geometry, prepare that CAD geometry, and then manually create weld geometries; after that, you weld the created geometry to the body you want to weld it to, and after that, you mesh it and troubleshoot problematic geometry. Only after that can you solve it and start Seamweld Analysis in Ncode. The New Weld function allows us to take these three actions and make them 3 in 1 using only Automatic Welds, and then after solving that, go straight to Seamweld Analysis.
The weld geometry that I was researching generates a weld using two planes and one edge. You can generate both intermittent and continuous seams. You can even change the properties of the weld, such as weld leg length, thickness, material, etc., to the desired one.
I created a time savings study with three different cases. The FEA analysis was invited to work on the manual welds, and I would do Automatic ones. In the best scenario for the Automatic Weld, time could be reduced by 87%, which would be a big deal for the team.
An average 2-3-tone lift truck frame has more than 150 welds, and it takes two weeks to weld it manually. I welded this frame in 3 days using an automatic weld function. This gives a 70% decrease in time.
This means that the Global Modeling & Simulation team can now apply Seamweld analysis to a much wider variety of projects, making it so much faster and easier. In addition to that, average lead time and productivity are increasing significantly since the team can now perform quick draft weld analysis to determine points of interest that will be manually welded.
I also finished the Automatic Welding Procedure, which includes Theory, Procedure, Limitations, Best Usage Cases, and Troubleshooting of commonly faced issues.
The next project will be an FEA for the C23 Wheel & Adapter and a P12 adapter for the Wheel Force Transducer. I did the simulation and verification of the proposed designs.
Let’s start with the C23 wheel. This is one of the most accurate simulations I did. I put all bolts, pre-loads, assumed base and all frictional contacts which made it a nonlinear simulation.
The results show stresses below endurance, which is the desired performance target. This means the part passes this validation.
The next one is the C23 Adapter. There was an issue with the radii value on this edge. It did not exist in the model but lived in the physical part. The instrumental lab measured it, and I created a 0.35 mm radius on it. That is a tiny radius; hence, the mesh must be even smaller to get accurate results.
Maximum principal stress is below the endurance limit of the material, so it passes this requirement, too.
The final FEA I want to talk about is the P12 Adapter.
The first design I received of the P12 adapter started to meet stress requirements after three revisions but was rejected due to high cost.
The second design was an assembly of two parts. I carefully simulated it, including all bolts, base, pre-load, and frictional contacts, which made it a Non-Linear Static Simulation and did not meet stress requirements.
I made FEA for the third design, too, but it didn't pass the maximum stress requirements.
The main goal of this project was to design and build an adjustable catapult to demonstrate and apply the principles of experimental design (DOE) in making a model based on the experimental data that will reliably predict what factors value combination will get us to the desired projectile hit distance. DOE, a structured and analytical approach that can be implemented to solve engineering problems, uses mathematical and statistical methods to efficiently and effectively collect data. When done correctly, these principles yield reliable, well-supported engineering conclusions while minimizing the number of data collection trials and saving time and resources. The catapult was designed with four independent variables, and each was engineered to influence the trajectory of the projectile, which, in this case, was a foam cow.
My part of the project was to work on the design of the experiment, data analysis, and model creation.
Design of Experiment
I decided that the full factorial model would be the best fit for the catapult design, with one of the upsides of using this methodology being the efficient use of time and money. Some of the variables that were previously considered, such as the angular displacement of the arm before launch and the distance of the bucket from the end of the arm, were not compatible with this model because they were continuous variables. Four independent control variables were required to be categorical and binary. Considering this, bands that could be added or removed would be the perfect variables because they can be absent (-1) or present (+1) in each trial.
After I decided on the independent variables, I started to design the experimental procedure. The blue band would always be present, so even if the two bands that served as independent variables were absent, there would not be an output of zero. To increase the accuracy of the results, I decided to increase the number of launches per setting to 3. I created three tables with 16 different configurations for a total of 48 launches. The model’s simplicity allowed the trials to be conducted in only a few hours.
I placed the catapult between the Engineering Building and the Fourth Avenue Building, where launch day was expected. All 48 launches were completed, and the catapult survived with no damage.
An AI randomly generated the initial order of the launches. The order of the launches was executed according to the columns below. At each step, I launched the projectile with the same setting three times. The idea was to make an order n+3. For example, I would use the fourth one after the first configuration.
Analysis of Experiment & Model Creation
The used model is a full factorial with four control variables that could only have one of two values, as seen in Table 1 below:
After I had collected the data, I decided to analyze it to see if there were any anomalies in the data. I decided that a 5% error in the data was an excellent goal and made a plot based on the raw data. The outcome of my plotting is that for Configuration 4 on launch 3, the output value exceeded my accepted ceiling of 5% error. Since only one out of 48 values did not meet my requirements, I decided to proceed with the data I had. The model showed great r and f values, and hence, I decided that the anomaly I have is not hurting my model in a significant way.
I decided that R would be used for data analysis, so the data was required to be in one big table. To do so, three tables with 16 launches each were merged into one table with 48. The first step in analysis using R was to import the table and review the data, as seen in the appendix. The second step was to execute the model and take a look at the first model.
The third step was to analyze the model and optimize it. Multiple variables were likely to have no impact on the model and, hence, must have been removed from the model. I removed insignificant variables one at a time. I will not show the complete iterative process due to its length, but the optimized final version can be seen below:
The fourth step was to verify and approve my final model. The confidence interval for all variables was not crossing 0, which was the desired result. All variables except for C:D were likely to have an impact, with chances higher than 99.9%. C:D interaction probability to have an impact was also high but only more than 99.77%p>
Conclusion
I created a model that accurately predicted the settings required to hit all three targets. The simplicity of the employed methodology allowed me to complete all necessary trials, and the resulting model predicted the launch distance with very high confidence. The settings I used to hit all three extra credit targets are shown below in Table 2: