"How do you automate aircraft assembly with robots?"
Aircraft assembly is complicated. Large lifts and cranes are neede to move large components, but they require a lot of manpower and time, because they must often be operated manually for complete control. To improve on this process, Boeing asked the NxR Robotics Lab at Northwestern to demonstrate the moving and orienting of large airplane components using collaborative robots working to manipulate a single object. Our task, as a senior capstone team, was to design a scalable prototype with robust engineering analysis to meet Boeing's specifications for force application, velocity, positional accuracy, and modularity.
Research and Early Mockups
We began with an extensive search into relevant IP and competitive products. We sought out systems that rival the XYZ linearly-actuated mobile base and lift systems that Boeing currently used. After identifying potential designs, we used an alternatives matrix to evaluate them against weighted factors, such as ease of motion, XYZ range of motion, modularity, and manufacturability. We chose a rotationally-driven delta design for their smooth translation, compact size, and easy integration with DC brushless motors.
Meeting the Specifications
In order to optimize the geometry of our delta to achieve the highest range of motion, we used Matlab to model the delta in 3D space. With the help of graduate ME students, we created a program that showed the 3D workspace of the delta given the lengths and angles of the arms. We later used this code to program the robot's motion to demonstrate basic motor control. Using G code, we input these angles from our Matlab code into FlashCut CNC software that then manipulated our stepper motors to the correct positions. The below graph shows a Matlab plot of the 2D workspace of the delta robot in meters.
Solving the Problem of Overconstraint
A major challenge of collorative robots is overconstraint. Unexpected external forces between multiple robots rigidly attached to the same article causes stresses in the article being moved. To introduce compliance into the system, we used a torsion spring between the output of the gearbox and each lower arm of the delta. This allowed approximately 5cm of xyz passive compliance at the end effector. We incorporated DFMA principles in order to reduce shaft misalignment and simplify the design of the spring module. For example, to reduce axial overconstraint, we chose to use a timing belt pulley to drive each arm instead of having the motor and gearbox in line with the arm. Below is the design of the spring module, which transferred torque from the gearbox to the arms of the delta robot.
Final Prototype and Future Work
At the end of the project, we developed an alpha-level prototype that provided linear actuation, passive compliance in all 6 degrees of freedom, mobility across a room, and control hardware for $5000. For comparison sake, the cost of a Fanuc Delta robot ranges between $25,000 and $50,000 depending on the model and size. We achieved a low cost, scalable robotic manipulator that provided a framework for the NxR Lab to iterate on for future parallel robotic designs. One suggestion our team had moving forward was to use custom machined torsion springs instead of stock torsional springs for increased accuracy and precision in returning the end effector to a home position.