The main objective of this R&D result is to realize a fully automated multi-sensor welding system, capable to adapt the welding parameters to the morphological condition existing along the joint, and react in time to every fluctuations which may happen during the process and to the disturbance variables.
When a weld is manually performed by a welder, this one takes appropriate actions by lifting or lowering the welding torch, or executing suitable swinging movements or again speeding up or fasting down the welding speed in accordance to the local condition of the joint to fill. In such a manner the molten metal deposition results smooth and regular compensating any irregularity of the joint.
Even if several automated or robotized solutions exist, they are not able to take into account the disturbance variables before mentioned. For instance, the control is only capable to reproduce the welding parameters previously pre-loaded in a database and choosing them in accordance to the “overall” geometry of the joint, and cannot overcome the fluctuations and irregularities associated to the specific morphological conditions of the joint. For example, the deposition of the molten metal can be influenced by irregularities in the joint profile or in the welding passes that have been already executed, like mechanical working marks, welding spatters or incisions.
These issues are particularly severe when a weld has to be performed in difficult situations, i.e. a welding position significantly different from a plane one or with a very curve profile, filling up a narrow-gap bevel in a thick joint and executing multi-pass welds. In those cases the execution of an automatic weld may result impossible.
This project aims to fill this gap by realizing a next-generation fully automated multi-sensor welding machine, driven by a novel self-learning control logic.