The ongoing miniaturization of electronic components is resulting in increasing requirements for solder paste printing, assembly and the soldering process within SMT production lines. Thanks to the large amounts of measurement data already available, models based on Machine Learning techniques lend themselves to improving the processes themselves, as well as the quality of the electronic assemblies. Further optimization potential lies in the reduction of downtimes of the individual systems.
The research project “Distributed Machine Learning in Electronics Production (D-LEAP)” deals with the utilization of already existing measurement data in order to solve these use cases with the help of ML models. Intelligent maintenance strategies, for example for cleaning the equipment, are to be developed. In addition, the models are to be used to make it possible to adjust process parameters across plants and lines in order to increase the quality of subsequent processes. Another aspect is the evaluation of the quality of manufactured assemblies. The majority of detected defects in electronics production are false calls. Therefore, assemblies detected as defective require further inspection. Also this inspection effort shall be reduced by the better predictions with the help of the developed ML-model.
Special attention shall be paid to the use of all inspection data (SPI, AOI, AXI) for the learning of the models, so that also defect origins can be determined as accurately as possible. Furthermore, linkages of the use cases will be investigated. For example, a decreasing quality of the assemblies can be used for a maintenance of the systems.