Senior Design Project

Project Pitch (2 Minute Summary Video)

Abstract

The senior design project of my Mechanical Engineering degree was to design a mechanical testing device for DuPont's Mobility and Materials division, which is now a part of Celanese. The device was to perform tensile creep tests on ISO standard specimens made from different materials, with key goals being to improve sample throughput and load consistency. Additional responsibilities included the development of a data collection program capable of measuring force, strain, and temperature data, with visualization and collection managed through a user interface. Also provided were specific criteria on desired loading rate, maximum load, and a temperature range for an included environmental chamber.

This project was completed in its entirety in about three months, from assignment to a project on August 31st until the Senior Design Showcase on December 14th.

Figure 1. ISO standard dogbone test specimens, the type to be tested by our device.

Figure 2. An environmental chamber, inside which a specimen would be placed.

Figure 3. Wedge grips with a specimen clamped in between.

Background

Creep is a material property that describes how a material plastically (irreversibly) deforms over time when subjected to a static, unchanging load. It is dependent on environmental conditions such as temperature and follows a theoretical triphasic curve, shown below. The timescale of these tests can span months to potentially years, which necessitates carefully controlled testing conditions. Maintaining these static conditions over long tests can make creep data difficult to obtain, despite the test itself being quite simple.

Figure 4. A sample creep curve showing the primary strain, steady-state creep, and tertiary creep phases of creep. The complete duration of the test shown is roughly 2 x 107 seconds, equivalent to about 231 days or 7.7 months.

Graphic adapted from Figure 13.3 of Materials: Engineering, Science, Process, and Design by Michael F. Ashby, Hugh Shercliff, and David Cebon.

To start, our group of four engineering students did benchmarking by examining some existing products on the market for creep testing. We considered the Instron 6800 Series, MTS Model C45 Wide, and Zwick-Roell LA DW and Spring models, which all varied by factors such as sample quantity, force loading method/capacity, and compatibility with an environmental chamber.

Figure 5. Instron 6800 Series

Figure 6. MTS Model C45 Wide

Figure 7. Zwick-Roell LA DW and Spring

Spring on left, DW (dead weight) on right.

Findings from these benchmarks and discussions with our sponsors at DuPont led us to the specifications of a physics-loaded, low-cost, compact and single specimen station that could be replicated and arranged in parallel.

As safety was a priority of ours, we needed a force amplification mechanism to reduce the required amount of input load and prevent injury to operators. Levers, block and tackle pulley systems, and gear trains were the methods of force amplification considered.

Concept Generation

Levers

The main advantages of levers are their very simple design (a rigid member with minimal moving parts) and thus low energy losses. For our application, the only useful work done by the lever is related to the force acting perpendicular to the horizontal (the force straining the specimen), a quantity which changes as the specimen deforms and the lever angle adjusts. Some calculations (shown under Derivation) were performed to quantify the changes in output (axial) force felt by the load train in terms of displacement and properties of the lever, arriving at the final expression:

F3 = F2 ∙ (L2/L1) ∙ [ cos(arcsin(∆x/L1)) ]2

where:

F3 is the output force

F2 is the input force

L1 is the distance along level from fulcrum to output

L2 is the distance along level from fulcrum to input

∆x is the vertical displacement of the lever, equal to the displacement of the load train.

The quantity arcsin(∆x/L1) is equal to the angle the lever makes with the horizontal.

Derivation

The derived relationship shows that in addition to the expected lever length factor (L2/L1), there is a factor whose value decreases as the sample displacement, ∆x, increases. When displacement is minimal, the resultant angle is small and thus cos θ ≈ 1 and [cos θ]2 ≈ [1]2 = 1. As the angle increases, the quantity begins to shrink increasingly fast.

As an example, a displacement of 1 centimeter (about 5.88% of total specimen length) causes a lever with a 10:1 distance ratio would have an amplification factor of only about 9.6, a loss of 4% in an ideal case.

Figure 8. Example calculation showing output force reduction due to sample displacement.

This makes testing of high-strain materials impractical without intervention. One method used by an existing system, the PLC Lever Arm by Applied Test Systems, is to incorporate a linear actuator to keep the lever level. This approach reintroduces reliance on electricity, which we intended to limit to data collection where it was absolutely necessary.

Figure 9. Portion of PLC Lever Arm manual's Figure C.2, with added red ellipse highlighting the auto-leveling actuator.

Figure 10. PLC Lever Arm manual's Figure D.1, showing the lever labeled in detail.

The options we saw to keep the lever level were to add physical alternatives to add a restoring force dependent on the angle, or to shift weight during operation. Using a linear spring to counteract quadratic decrease seemed a flawed solution, and by the time we considered weight shifting concepts like weights on rails or viscous fluids and valves the level had lost most of the simplicity that had made it so appealing at the start. This leveling problem, along with concerns about the compactness of a long lever arm, led us to investigate the next amplification concept:

Block and Tackle

One thing that made a block and tackle force amplification system attractive for our use was that all forces are exerted through the axis of the cable, which resolved lever related concerns about force changing as the specimen strains. The ability to add more sheaves to the blocks or create a chain of suspended pulleys allows for a greater amplification factor in a relatively compact space. A consequence of a block and tackle system, however, is greater friction losses between the many wheels. Though difficult to find, we were able to find one reasonably sized option for an eight sheave block and tackle as stock hardware online.

A new problem that became apparent was that the travel distance of the suspended weight would be equal to the mechanical advantage multiplied by the specimen's displacement. This meant that to maximize potential strain of the specimen, the weights should begin as far off the floor as possible, which would require an additional system. Setting that problem aside for later, we quantified the possible testable strain in a frictionless ideal case, considering a weight travel distance of one meter.

Figure 11. Calculations showing displacement of the weight stack in an ideal case (top) and the amount of strain a sample could experience in an ideal case if the height change of the dead weight were restricted to one meter (bottom).

Another consideration of block and tackles is the need to select a cable with enough strength to safely lift the applied load. Stiff and strong braided steel cables do not bend sharply around small pulleys, while flexible fiber cables are more compliant and would appreciably strain along with the specimen. An ideal cable material would have both a small bending radius and high resistance to strain, but using the available stock block and tackle would require a tradeoff of flexibility at the expense of stiffness. Thinking about ways to circumvent this need for both flexibility and stiffness led us to the final concept:

Gear Train

Advantages of a gear train for our application were most similar to those of a block and tackle, with forces being applied tangentially to the gears (and thus axially through the load train if arranged properly). A sprocket and chain system could provide the balance between cable stiffness and small bending radius that we desired. A gear train also had potential for high mechanical advantage in a compact space, depending on the teeth of input and output gears.

With new benefits come different problems, though. Among our concerns were the ability of such trains to reserve ample chain to handle large strains. Great force amplification could be achieved if the input gear had few teeth, but it would also revolve many more times as a result and would need much more chain. We were unsure where this chain would reside at the start of a test, since in most cases involving gears and chains the system forms a closed loop with chain feeding back into the system.

We began considering wheel and axle alternatives placed adjacent to a gear instead, which led us back to the stiffness vs. small bending radius issue. A gear train might also demand tensioners to keep the load train taught as the sample strains, as well as a housing for the gears to protect operators from pinch points. Finally, the gear train concept did not relieve us of the need to start weights far above the ground. These additional complexities made gear trains less attractive in comparison to the simpler block and tackle.

Figure 13. An alternative idea with gear rotation being driven by a wheel connected to a common axle.

Figure 12. Early pulley belt and gear train concepts. The ends of input and output cables as depicted would not allow for significant strain of the load train.

Concept Selection

After doing research into these three amplification methods, our team judged them using a weighted matrix with five criteria: Reliability, Mechanical Advantage, Simplicity, Compactness, and Energy Losses.

Reliability was given the half of the overall weight since creep tests are extremely time consuming to perform, and the loss of applied force would render the test results unusable. We also wanted the machine to be a "set and forget" operation that would not demand frequent check-ins over the test length to confirm everything was working properly. While reliability is a broad topic that is related to other criteria, we used it to gauge each concept's innate potential for error in the applied load over time.

Mechanical Advantage had the next greatest weight, since greater mechanical advantages reduced the amount of weight required, thus making the machine safer and easier to load. Greater mechanical advantage also allowed for a wider range of possible loads and possible data to generate.

Simplicity came next. Though simplicity and reliability are related to one another, our focus in this criterion was primarily on the ease to operate the machine and troubleshoot any problems.

Compactness was related to our goal of placing many of these machines in parallel. The smaller the volume we could contain the system in, the more machines could fit in a room and the more data that could be generated in the same amount of time.

Energy Losses was the last criterion considered. Similar to mechanical advantage, concepts with less energy loss would require less input weight. We mainly considered losses due to friction, as they were our greatest concern.

Table 1.  Weighted matrix for concept selection. Concepts were assigned scores from 1 (lowest) to 5 (highest).

The results of our selection led us to settle on a block and tackle system to handle force amplification. As a quick spot-check, our team assembled a small prototype using four pulleys to confirm the concept was feasible.

Figure 14. Small scale prototype to confirm the feasibility of the block and tackle concept. A piece of elastic held in place by a pair of clamps stretched as the gray paracord was pulled and the carriage translated upwards.

To handle elevating the weights, we produced a single concept utilizing a remote controlled linear actuator. When retracted, weights would be loaded onto a cradle placed onto a plate at the actuator tip. The actuator would then be extended, the cradle connected to the block and tackle system via carabiners, and the actuator would retract leaving the weight suspended.

Figure 15. Linear actuator weight release concept. A flat plate on the actuator tip supports the weight cradle during loading and unloading. Weights (shown in red) are threaded onto a cradle (in white) connected to the block and tackle cable (in yellow). The linear actuator (in blue) rests in a custom mating seat connected to a wide plate to discourage tipping.

When selecting sensors, we considered the following:

Force

Displacement

Temperature

After discussions with our faculty advisor and DuPont about the hardware available from their storage as well as what we were capable of accomplishing in our three-month development cycle, we decided to use an OMEGA LC103B-3K S beam load cell, Epsilon 3452 clip-on extensometer, and a OMEGA LD620 LVDT. The scope of the project was narrowed to eliminate the environment chamber due to time concerns, though efforts were made to design the load train to accommodate the addition of a chamber in the future.

At this point our team began delegating tasks between members, though we all helped each other as needed and worked on common tasks like presentations and reports. My teammates being primarily responsible for the frame, load train, and linear actuator assembly; my primary responsibility was to write the code for the data acquisition system.

Prototyping

Data Acquisition

Having worked with the LabVIEW software on a data collection project during my biomechanics research, I had a good foundational understanding of how to reach the outcomes we needed. Over the course of the following weeks, I met with our contacts at DuPont to learn proper configuration of the various sensors and clarify desired capabilities of the program. To minimize file size over the long test duration, I added logic to record a new data point if and only if the extensometer read a change in strain greater than a threshold value, specified by the operator at the start of the test. This threshold and other parameters of the test are saved to a text file, while sensor readings are saved to a delimited file to be imported into analysis software. Also accounted for by the program were current absolute sensor readings, current test duration, collection of no-load values to obtain scaled sensor readings, and graphs for all sensors for visual feedback, among other things.

Figure 16. Example of an output text file containing test parameters.

Figure 17. Test data file. By column from left to right, the data is Time(s), strain, force (lbf), and LVDT displacement (mm). As seen from rows 56 onward, the time between data points increases dramatically while the strain between data points increases at a nearly constant rate.

Figure 18. Front panel for the data acquisition program. Operator inputs are entered on the left, test control and current readings are displayed in the center, sensor graphs are along the bottom, and debug information is shown on the right. The assistance feature in the lower right was intended to use curve fit data from a series of tests to compensate for losses in the system and guide operators to what input weight would yield an output load. Having gotten the code in a workable state for testing, I joined the rest of my team in prototyping.

Frame Design

To get a sense of our frame design, we constructed a wooden mock frame to approximate the actual size of the eventual final prototype. From the mock version, we lowered the height of the grip platform and determined the quantity and lengths of material to order. Based on the expected deflection of aluminum under the desired maximum loads, we decided to use 80/20 framing hardware for a modular and easy to assemble frame.

Figure 19. Wooden mock frame intended to match the size of the final prototype. This frame was helpful in determining that the grip collar needed to be lowered, and in selecting lengths of material.

Final Prototype and Deliverables

The next month was spent ordering stock hardware, machining any necessary custom pieces, assembling systems, and testing the code. I designed and machined a pair of LVDT mounts and a pulley mount to both be compatible with 80/20, produced a user manual for the project, and performed validation testing of our project metrics.

Figure 20. Assembly drawing for an LVDT mount.

Figure 21. Assembly drawing for a pulley mount.

Figure 22. Machined LVDT mount in use.

Figure 23. Machined pulley mount in use.

Figure 24. Image before my retouching.

Figure 25. Image after my retouching.

The final prototype frame was made from a machined plate connected to an 80/20 frame. Adjustable feet allow for leveling of the frame on uneven surfaces. All frame hardware could be assembled by hand with a single hex wrench.

Figure 26. Final frame, with no load train, rails, or suspended weight installed.

To improve the safety of the device, a drop pad assembly was created. High-density foam was taken from weight crash pads and placed inside a box with ABS plastic walls. If the cable were to fail during a test, the drop pad would catch and contain any falling weight to prevent harm to the frame or any operators.

Figure 27. (Top) Drop pad, empty and with weight inside. (Bottom) Weight being released into the drop pad.

Figure 28. Frame with suspended weight and drop pad in place below.

The final load train consisted of a load cell fastened to both the machined plate and the stationary lower grip, the specimen clamped between a pair of grips, a movable upper grip connected to the underside of a carriage, and the block and tackle connected to the top of the carriage. Motion of the carriage was constrained to translation in a vertical plane through a pair of rails.

Figure 29. Labeled diagram of the load train.

Figure 30. Testing the load train by tugging the cable with a rubber specimen.

Figure 31. Loading weight onto the cradle, and using the linear actuator to raise and lower it.

Figure 32. (Left) Components of the weight release assembly. The tip plate underside has an aluminum tube welded in place, sized to snugly slip over the actuator tip. The top of the tip plate has a rubberized surface to stabilize the weight cradle. The base plate is custom machined to fit the model of linear actuator used. (Right) Assembled weight release system.

Figure 33. Retracting the linear actuator after connecting the weight cradle to the block and tackle via carabiner. The weights remain suspended and place the block and tackle into tension.

Validation and Conclusions

With the weight release linear actuator working, we could begin design validation. We did not have weeks to run a proper creep test, so we set aside 18 hours to see if any glaring issues arose. From the data we were able to collect, it was clear that compliance in the block and tackle's fiber cable was reducing the effective load experienced by the specimen over the duration of the test. This compliance also led to issues where the linear actuator could not retract enough to suspend the weight cradle.

Figure 34. Front panel view at the end of the 18 hour test. The force graph in the middle shows a downward linear slope, indicating the force applied through the load train was not constant as necessary for creep testing.

Figure 35. Straining of the test specimen when compared to a control specimen.

Our suggested path forward for this project included four main goals:

Though our final prototype was not a market-ready product, we were confident our work over three months showed the concept had potential worth exploring in greater depth.

At the design showcase, our team was honored to receive the Professor Michael Keefe Award, an award named after a University of Delaware professor given to a senior design project that best embodies design iteration and the engineering design process.

Figure 36. Evolution of the machine from a small scale proof of concept, to a full scale mockup, to the final functional prototype.

Figure 37. Fellow team member Oscar Marino and I after receiving the Prof. Michael Keefe Award for our work.