An Approach to Monitoring and Guiding Manual Assembly Processes

With the dawn of Industry 4.0, companies are constantly looking for ways to digitize their work flows in order to quantify, monitor, and improve work in industrial settings. While some processes, such as those using computer numerical control machinery, are either easy to embed sensors into or already have sensors embedded that are providing information, there still are core processes taking place in industrial settings that do not have easy ways to retrieve data from. One such process is the manual assembly process. In addition to the difficulty in retrieving data from assembly processes, the assemblies that rely on human operation with simple tools rather than machining centers are more difficult to immerse in Industry 4.0 practices due to its primarily offline nature. To better address this issue, we propose an approach and describe implementation of the corresponding system that is capable of providing information to assemblers based on data received during an active process. Specifically, the system informs the assembler of the next steps in the assembly process based on what the platform determines has already been done. In order to accomplish this, a convolutional neural network is used to analyze images on a camera that will be overseeing the physical assembly process. In this way, there will be a live digital record of the assembly process taking place as well as some, albeit basic, interaction between the assembler and the overarching system monitoring the process.