Published papers

This study presents a video analysis method for assembly factories. In recent years, there has been an increasing demand to collect manufacturing data in the manufacturing industry. While visualizing manufacturing data in real time to analyze practical manual work has not yet been realized, technologies for data collection, analysis and visualization have made progress using many kinds of machines and sensors. The introduction of various sensors and machines has resulted in changes in processes for visualizing the collection of manufacturing data in real time and analyzing it, as well as the collection and analysis of data. While difficult to do manually, the real-time monitoring of production data can help to quickly respond to problems, investigate causes, and improve operations. Therefore the authors propose a work analysis method that uses videos to analyze hand movement and position and product position. As a result, an analysis accuracy of 99.3% is obtained when studying standardized work in an ideal capture environment, and the cause of the reduction in accuracy is clarified.

Gas burning that occurs during the resin molding process is fundamentally due to the improper venting of gas. However, finding a solution is extremely difficult as the phenomenon arises from a complex interplay of numerous factors. In the maintenance of molding machines and molds, carrying out regular inspections, replacing parts, cleaning, and maintaining the machine condition are practical and effective measures. During quality control, it is important to perform maintenance at the appropriate timing, but the timing is often determined based on experience and intuition, meaning that there is a lack of clear standards. This paper proposes a method using a deep learning classification model to clarify the timing at which maintenance should be performed based on the perspectives of quality and efficiency. In the proposed method, first, the cosine similarity between images of an initially molded product, which serves as a standard, and images of each subsequent product is calculated. Then, transfer learning of the deep learning model is performed in which classification is performed based on similarity. Finally, fine-tuning is performed. Emphasis is placed on the utilization of a loss function that directly controls the feature values during the fine-tuning.

In high-mix, low-volume production, workers are required to acquire skills for various work procedures, standardization of work and promotion of proficiency are important challenges. This study develops a work analysis / management system using digital technologies to monitor assembly work in real time, collecting and analyzing daily work log data to visualize work progress, assembly time for each part and variations in assembly sequence. The system enables a detailed understanding of the current state of work, which was previously difficult to analyze, and supports identify work that requires improvement. The system also enables workers to evaluate differences in skill levels and variations in assembly work. As a result of the implementation of the system at actual production sites, it is shown that the findings can be used as new information that contributes to clarifying priorities for work improvement, standardization of work, and promotion of work proficiency. This study proposed the developed system as an effective work analysis / management method for high-mix, low-volume production, and verified its usefulness at actual production sites empirically.

Production line design is one of the important research topics for constructing smart factories. Many research papers have been published concerning smart factories; however, most of them are focus on subjects such as automation, IoT (Internet of Things), AI (Artificial Intelligence) and other new technologies that are introduced to increase the efficiency of existing manufacturing lines based on the so-called philosophy of “ status-quo improvement ”. In this research, we consider a design problem for a new production line taking into consideration investment cost, operation cost, synchronization penalty and flexibility penalty. We emphasize synchronization and flexibility as new key performance indices to evaluate the level of intelligenced to the smart factory, in which a multiple number of equipment in two production processes can provide the proper number of combinations to satisfy the external demand remove. Based on a real case involving a major player in the dairy product processing industry in Japan, we focus on two core production processes in a processed cheese manufacturing line; melting process and filling process. We design the process to include investing in a multiple number of machines with different capacities, while minimizing the total cost giving consideration to investment, operations, synchronization and flexibility. Two types of decision variables are introduced representing investment and operation. Through numerical experiments and the case study, we show that the proper number of equipment and capacity can be obtained applying the proposed model. The proposed model can also be applied to those companies considering further investment in factory or production line.