6-DOF Manipulator: Previous Year’s Design

During the 2018-19 competition, the arm performed better than expected, but not nearly as well as I know it is capable of doing. Being at competition, the shortcomings were quite obvious. In any task that required significant arm interaction, too much time was spent wrestling with the inability to intuitively control the end effector or managing movements based on the resulting “slosh” of the arm from inertial loads. At the time, the mechanical design of the arm was the blocker in implementing  a proper control scheme. Some of the mechanical issues which contributed to this shortcoming are outlined below.

 

• The vast majority of the ANSYS simulations from previous years were either improperly setup (Boundary conditions, force magnitudes and directions, timesteps, contacts) and in turn do not reflect reality in any way, do not model the material properties of composites properly, or do not seek to obtain metrics that are now deemed necessary to gauge the performance of the arm. 

• Excessive system backlash and lack of manipulability: You should not be able to grab the end effector of a what is supposed to be a precision robot, and move to 30 degrees in all directions solely from the play between mating components.

• • Improperly constrained gears

• • Excessive shaft to bearing clearance

• • Improperly torqued fasteners 

• • Bearing to bearing bore misalignment leading to rapid wear

• • Poorly manufactured transmission shafts 

• Non-intuitive controls lead to difficult interacting with competition tasks: The overall design and construction of the arm was far below the required threshold to implement and reliability trust an inverse-kinematics control scheme. 

• Zero inter-assembly design integration: Each joint was designed independent of one another and they were all kind of hodge-podge connected a couple of days before the end of the semester. The arm lacked any real cohesion. The connections became poorly thought out weak points as it was never clear which half was responsible for the design, and things were done very last minute.

• Over-engineering from a strength of materials standpoint

• • ANSYS FOS of 15 on many components

• • Unnecessarily heavy, reducing overall mass margins

• • Use of difficult to analyze link materials with very little available technical data

• Difficult to manufacture and assemble: It is important to draw a distinction between two types of “difficult to assemble” parts. There are parts that may require jigs to assemble, or a prescribed process, but are straight forward to assemble. There are also parts and assemblies whose assembly process is inherently difficult due to the complexity of the assembly, and any list of simplifying possibilities have been exhausted. 

 

Then there are parts and assemblies that are relatively simple or  that do not inherently require a difficult assembly process. Parts in the past were designed in such a way that they literally could not be assembled because another part ends up blocking the placement of a fastener or tool. This also tie’s back into the DFM vs functionality discussion. The joints in the past were made from several plates all fastened together. This appears to be the “easy” way to manufacture just because no CNCing is involved, but it not only makes it more timely to manufacture and change if needed, it makes it more difficult to manufacture and assemble (leading to the aforementioned condition), makes it more difficult to analyze, and almost certainly less robust. A further explanation is given below: 

 

• • Making joints from several plates means more individual features to assemble the plates.

• • Ensuring bearing bores are concentric is more difficult as a single datum structure is not being referenced.

• • Necessary to machine, keep track of, and assemble several parts instead of a single part, unibody style construction.

• • No one on CMR is a true machinist, and the CNC will produce more accurate parts in hours as opposed to days.

• • If a design change is required, it’s a matter of updating the CAD and because the CAM is associative, we can just re-run a one hour program instead of machining new parts over a several day span.

 

I recognize that the previous years method of manufacture would likely be cheaper and faster in industry. However, the infrastructure required to make this process a several second process instead of several days would incur a significant upfront expense. We are also not driven like industry, our part count is relatively small, and we shoot to achieve the most functional product, not the cheapest.