During the welding process, input factors (welding current, welding cycle, instantaneous heat rate, welding force, electrode tip diameter, etc.) all affect the quality of the welded button created by a resistance spot welding machine (RSW). Fixed input parameters such as welding force and welding cycle can be controlled and adjusted manually based on welding standards, but variable input parameters such as welding current and instantaneous heating rate (IHR). IHR and electrode tip size cannot be controlled manually. This study shows the relationship between welding current, heating rate and electrode tip size affecting welding quality. Classifying weld quality and adjusting appropriate welding current using an artificial intelligence model is proposed in this study. An artificial intelligence (LSTM – Long-short Term Memory) model is designed with several input parameters of the RSW machine. For RSW machine using an AC inverter, the quality of the weld depends on various influencing factors such as welding cycle, compression force, electrode cross-sectional area, welding current, etc. This paper is based on the calculation of IHR values by a data collection software program. Based on the IHR values, the weld quality management system can classify the quality of the weld at the current moment. To predict the weld quality, the correlation between IHR values, electrode cross-sectional area, and welding current is analyzed and evaluated. The performance evaluation indicates that the LSTM model in predicting the trend of IHR value with training data, namely MSE, MAE, and Coefficient of determination (R2), are 0.5, 0.25, and 0.78, respectively. For validation, the evaluation of LSMT's performance is 0.8 for MSE, 0.64 for MAE, and 0.8 for R2, respectively.

This work develops a financial distress prediction model by combining a multi-layer perceptron artificial neural network (MLP-ANN) with the conventional Altman Z-Score model. Financial data comprising 7,126 observations from 509 listed firms in Vietnam, collected between 2006 and 2019, were used to build the hybrid model. Our predictive model accurately classifies the financial status of listed firms into three categories: low bankruptcy probability, high bankruptcy probability, and a gray region of bankruptcy, with an accuracy rate of 98.4%. Additionally, the results segmented by industry show that the trading company model achieves the highest accuracy (99.2%) in identifying the firm's financial status, followed by the models for the service and manufacturing industries (98.6% and 97.4%, respectively). In Vietnam, the Z-score and ANN hybrid model proves to be both appropriate and useful. Banks, securities firms, and investors can utilize this application to forecast the financial status of businesses in advance, aiding in making well-informed decisions. CCS CONCEPTS • Information systems • Information systems applications • Decision support systems • Data analytics