Machine-learning-enabled Decision Support for Screwdriving Process

Screwdriving as one of the most typical processes in production is facing massive amounts of data. These data empower engineers, managers, and manufacturers to understand the screwing process and make decisions. However, the prerequisites are the ability to turn the abundance of data into knowledge and the intensive cooperation between humans and machines. Nowadays, technologies like big data analysis, machine learning (ML), and artificial intelligence (AI) can be integrated into Cyber-Physical Production Systems (CPPS). They help to get a better insight into the manufacturing processes. However, utilizing these techniques can require human efforts and expertise. In this paper, we show that machine learning can support decision-making in the screwdriving process. The proposed approach is validated on an industrial-scale screwdriving case study in a learning factory. Under the CPPS framework, a data analysis and machine learning toolbox is developed. Our test results show that this toolbox can help users understand the screwdriving process with less expert knowledge. The random forest algorithm as the best fit could effectively identify the screwing condition with an accuracy of 0.93 and F1 score of 0.90.