Classification of battery slurry by flow signal processing via echo state network model

In this paper, we propose a novel method to classify battery slurries using echo state network (ESN) model with real-time pressure and flow rate signals during circulating channel flows. To collect the signal, a closed circuit flow system with a pump, pressure sensors, and flow rate sensors is installed. The slurries with different states are prepared by two methods: long-term circulation and dispersant content control. Sensor signals are collected while the slurries are flowing through the pipe system. The collected signals show distinctive chaotic fluctuating patterns for different slurries, which are assumed to reflect the states of the slurries. The hidden state of the ESN is generated from these collected data, which are then split into training and test data. Consequently, the ESN can effectively distinguish the slurries by the output (label). We also analyze the accuracy of the network, based on training time and output averaging time. This study demonstrates that the states of the slurries can be detected from the fluctuating flow signals. We argue that the manufacturing process of any complex fluid can be optimized with this approach.