Agitation and mixing processes automation using current sensing and reinforcement learning

Abstract Many agitation and mixing processes utilize various sensors for real-time monitoring and control, which can involve complex and costly equipment. For many mixing and agitation processes, such as in dough making, as mixing energy is placed, the resistance to extension increases and then after some point it decreases again. High-quality bread is obtained by stopping mixing at or close to the maximum resistance. The change in resistance causes a change in motor torque. The torque change affects the motor’s current draw for agitation and mixing machines driven by electrical motors. The rheological characteristics of the mixed material are related to motor torque of the mixing machine. Therefore, it is related to the motor electric current where the load variation can be estimated by a low-cost current sensor. This paper outlines a novel design for an intelligent agitator/mixer process controller. The design is based on current sensing and on-line learning through reinforcement learning using operator input. The system provides a low-cost approach to automate various kinds of production equipment currently operated manually, which are common in the developing world. Additionally, the approach requires minimal modification to the equipment: it requires only a current sensor, an on/off control relay, a set of buttons for operation, and an embedded system.