Modeling and Control of Roll-to-roll Winding Machine in Li-ion Battery Manufacturing

Abstract

This work aims to investigate modeling and control problems of Roll-to-roll (R2R) winding machine. Tension control and lateral position control problems of a winding machine are the main research focus, a poor control of these leads to safety issues of the batteries. Novel modeling and control schemes are raised to solve the problems, the largest contribution is the proposed methods are highly feasible for industrial implementations. The feasibility of the proposed methods is supported by simulations and experiments. All experiment data is collected from a real industrial-scale winding machine prototype. In Chapter II, a lightweight sliding mode control-fuzzy logic control (SMC-FLC) algorithm is designed to solve the tension control problem. When traditional PID control fails to achieve satisfactory control performance, new methods should be developed. There are many advanced and complex existing studies regarding tension control, they achieve very satisfactory results in simulations and lab-scale experiments, however, most of them do not concern about saving the development and maintenance cost and the feasibility of generalization to large-scale production, yet this is one of the most important topics in real industry. The development and maintenance cost of the proposed method is less compared to the PID control because it is an adaptive and self-tuning algorithm, thus, it is an economical and highly flexible choice for tension control in large-scale industrial production. Being light-weight and simple, the performance is limited when higher operation speed is required, hence, a more advanced tension controller needs to be developed. In Chapter III, such an advanced model-based optimizer is designed to solve the tension control problem for a greater accuracy and faster convergence. There is a trade-off between easy implementation and greater performance, thus, to reduce the tuning and maintenance cost, the advanced control is designed to follow the iterative learning control manner, that is, allowing the controller to learn the control law adaptively. In general, Chapter II and III solve tension control problem with different focus: Chapter II focuses on easy implementation while Chapter III focuses on greater control performance. In Chapter IV, the calendering lateral position control problem is addressed. A R2R calendering device on a winding machine introduces a lateral position deviation problem when calendering the webs. Existing works either has large modeling error for a calendaring system’s web lateral position dynamics, or do not focus on web lateral position control. In this chapter, a hybrid physics-based and data-driven modeling method, together with a corresponding model based controller, are proposed to address the web lateral deviation problem. The proposed strategy is shown to be superior compared to the benchmark, and the method is generalizable to similar calendering scenarios. Conclusions and future works are discussed. There are certain limitations in this research requiring further improvements. Readers can find the suggested future study directions.