Strict industry demands necessitate better penetration classification in gas tungsten arc welding (GTAW) of aluminum alloys. An innovative classification method, namely time-frequency spectrogram based convolutional neural network (TF-CNN) has been developed in this paper. The logarithmic time-frequency spectrograms are used to interpret the raw arc sound data in the time-frequency domain and served as the input of an optimized convolutional neural network (CNN) model for welding quality classification. The L-ReLU activation function and RMSprop optimizer are utilized to improve the property of CNN. Effective arc sound features are extracted automatically by the CNN model. The proposed method is experimentally verified to recognize four penetration states with an average accuracy of 98.2%, which significantly outperformed four typical comparison models. The generation mechanism and characteristics of arc sound are discussed as is the adaptive feature extraction of TF-CNN. These studies expand the application scope of CNN to intelligent welding and achieve notable effects.

Most aluminum alloys, especially the 7xxx series, are incompatible with laser additive manufacturing due to many factors. In-situ process monitoring technology offers an excellent opportunity for process understanding and control, which improves the component quality and repeatability of AM. In this study, emission spectroscopy and plume imaging were utilized to monitor and understand the laser-directed energy deposition of Al7075 alloy. Characteristics of plasma were analyzed using spectra and plume image signals. Emission lines of Al I, Mg I, and Cr I observed in spectra signals demonstrated that primary elements in Al7075 were evaporated and excited during the AM process forming a plasma plume. The plasma plume intensity increased significantly as the laser power increasing. The periodic fluctuation of plasma was revealed using plume images, indicating that laser powers inputted into the target surface were unstable due to the laser and plasma interaction. The elemental evaporation and inconsistent laser input are the main challenges for Al7075 AM and are potentially monitored by plasma signals. Methods of signal process and feature extraction for spectra and plume images were developed. Experimental results showed that the aluminum emission spectral energy, the average baseline spectral intensity, and the plume area are practical for rough surface monitoring. Correlations between spectral features and key printing parameters were developed, demonstrating an excellent response of spectra features to the laser power and laser scanning speed. This investigation revealed the potentials and challenges of emission spectroscopy in the in-situ monitoring of laser AM, making progress towards intelligent additive manufacturing.

Owing to the high optical reflectivity of copper powder, the high-performance fabrication of copper alloys in the laser additive manufacturing (AM) field is problematic. To tackle this issue, this study employs the remelting process during laser powder bed fusion AM to fabricate defect-free and high-performance CuCrZr alloy. Compared to the non-remelting process, the remelting process yields finer grains, smaller precipitates, denser dislocations, and smaller dislocation cells. It realizes not only the dense molding of high laser reflectivity powders but also excellent mechanical properties and electrical conductivity (with an ultimate tensile strength of 329 MPa and conductivity of 96% IACS) without post-heat treatment. Furthermore, this study elucidates the influence of complex thermal gradients and multiple thermal cycles on the manufacturing process under the remelting process, as well as the internal mechanisms of microstructure evolution and performance improvement.