Finite element simulation of collisions between cast metal components on industrial robots

Sammanfattning

Industrial robots are an essential part of manufacturing products in modern industries through their ability to efficiently perform repetitive or dangerous tasks without human interaction. In the case of articulated industrial robots who are able to carry large and heavy loads, often for the application of automobile manufacturing, the risks associated with the robot colliding with its environment become important. While such collisions are sought to be avoided, understanding what could happen in a worst case scenario is still useful when considering safety in regards to the robot. One such worst case scenario that is investigated in this thesis is the metal on metal collision that could occur if the rubber mechanical stops are not replaced after a previous collision. The rubber mechanical stops permanently deform after a collision from the arm of the industrial robot rotating around an axis more than it is allowed thus causing a collision with itself. To study the structural behaviour from a collision on the robot arm, the theory of contact mechanics is investigated in combination with finite element method calculations. The collision is modelled using multiple methods which are then compared to each other. The first two methods compute a contact force which is then used in a finite element calculation to determine the stress state. The other method investigated explicitly models the contact within finite element software and thus determines the stress state directly. To verify the methods, one of the contact force formulations is compared to experimental data from previous research. Applying these methods to a simplified geometry of a robot arm provides results that are very varied in terms of both contact force and Von-Mises stress. While the contact force formulation that was verified against experimental data did agree relatively well in regards to the permanent deformation, the contact force was much lower than expected. Since the methods provide a large range of values, it can not be concluded which one is a best representation of reality, only their relative performance to each other.