Although laser vision systems for the automation of complicated arc welding processes have been available as commercial products since the mid 1980s, industrial breakthrough has not yet taken place. Up to now unreliability and high prices preventpotential users from investing in this innovative technology.The reliability and accuracy of triangulation based laser stripe sensors are significantly reduced by the characteristics of the open welding arc such as bright light, spatter and fumes that distort the sensor's camera signal. First steps to reduce theseimpacts are mechanical shielding, optical bandpass filters and increased laser power. The paper's first part discusses the results of experimental investigations carried out to quantify the efficiency of these physical measures. Although a significantreduction can be obtained, these physical measures do not eliminate all distortions caused by the welding arc. Therefore new algorithms to filter out the remaining distortions are discussed. Improvements are also quantified using experimental data.Finally new hardware platforms to realize these new algorithms are investigated. Powerful solutions such as parallel-DSP-boards or transputers should be avoided as they are too expensive. Therefore the suitability of conventional PC-processors and FPGAs(Field Programmable Gate Arrays) is investigated.
展开▼
