Construction of Motion Mode Switching System by Machine Learning for Peristaltic Mixing Conveyor Based on Intestinal Movement

Abstract

The high frequency of rocket launches requires low-cost solid rocket fuel. Currently, the fuel manufacturing process faces increased launch costs caused by the risk of ignition from rotary mixers and increased equipment and labor costs from batch processes in which mixing and conveying are separated. Therefore, this paper proposes and verifies an automatic switching system between mixing and conveying modes for a peristaltic mixing conveyor that enables safe and continuous mixing and conveying of solid fuel. In a previous study, peristaltic mixing conveyor with low shear force was developed and successfully produced solid fuel. However, there was room for improvement for more efficient fuel production because the device was controlled by pre-determined driving pattern. The actual intestine generates movement autonomously by enteric nerves. Therefore, the development of a sensing function that imitates the enteric nervous system and generates movement patterns based on the acquired data is expected to improve manufacturing efficiency. In this study, the sensor data of a mixed solid fuel simulant packaged in a bag were acquired, and the degree of mixing (unmixed and mixed completely) was discriminated using supervised learning (the k-nearest neighbor method). Furthermore, a system was constructed to continue the mixing mode when unmixing and automatically switch the motion to the conveying mode when the mixing was complete. The experiment showed that the motion mode automatically switched to the conveying mode at almost the same time as the labeled training data, and mixing and conveying of the simulated material was successfully performed.

Keywords

• Solid fuel
• soft robotics
• robot sensing system
• machine learning
• predictive
• data acquisition