Welding system with adaptive algorithm

Abstract

A welding system comprises a welding actuator (101) operative in accordance with actuator control data, a sensor system (102) for producing sensor data related to a welded seam, and a processing system (103) for determining the actuator control data with an adaptive algorithm on the basis of the sensor data and welding data expressing material to be welded and welding conditions. The welding system comprises a training data interface (104, 105) enabling a user to input the actuator control data for a training welding process. The processing system is configured to train the adaptive algorithm in accordance with the welding data, the actuator control data inputted via the training data interface, and the sensor data measured during the training welding process. Thus, the welding system can be trained so that the user is in the control loop. The adaptive algorithm can be based on for example an artificial neural network.

Description

technical field [0001] The present disclosure generally relates to automated, mechanized and / or robotic welding. More specifically, the present disclosure relates to a welding system, to a method for training a welding system, and to a computer program for training a welding system. Background technique [0002] Welding is a common method for permanently joining components of mechanical structures such as eg ships, railway installations, building structures, boilers, pipelines, power stations, aircraft, automobiles and the like. Although welding has many benefits, however, the joint formed by welding is often the weakest area of ​​a mechanical structure. In particular, imperfections in welded joints are problematic because they are potential starting points for mechanical failure. As a corollary, in most cases re-soldering is required when a weld defect is observed. Thus, welding defects increase both cost and production time. In companies making welded products, the cos...

AI Technical Summary

Problems solved by technology

Due to the nonlinear nature of the welding process, it is difficult to develop a model for predicting the behavior of the welding process

Therefore, it is challenging to configure the processing system described above to be able to determine welding parameters and possibly other actuator control data such that adequate welding quality can be achieved under different welding conditions

Images