Desenvolvimento de um controlador PID na plataforma de prototipagem Arduino para controle de vazão de um tanque

Abstract

Control and automation systems are crucial in the operation and competitiveness of various chemical industries. These systems provide means for monitoring, regulation, and optimization of industrial processes through the collection and manipulation of process variables, making the production process more predictable, consistent, safe, and less prone to failures. Among these control systems, the PID controller (Proportional, Integral, and Derivative) is the most commonly used for dynamic systems control. This system relies on the combination of three adjustable terms that act on the error value (difference between a measured value and a desired value) through adjustable coefficients to generate a control action on a manipulated variable, which is sent to the system to adjust one or more controlled variables. Proper adjustment of these coefficients is crucial to achieve optimal PID controller performance with quick response, stability, and suppression of unwanted oscillations. However, complex control systems are costly and require high process adaptation to accommodate them. Based on this, the present work aimed at implementing a flow PID control system, whose algorithm was implemented on the Arduino prototyping platform. The control strategy used aimed at controlling the feed flow of an electrolytic system by manipulating the continuous voltage of a pump. This implementation was carried out in three parts: initially, a supervisory system with a userfriendly graphical interface was built using Python to enable monitoring of system variables and operator interaction; then the dynamics between voltage and flow were analyzed through direct manipulation of voltage and recording of flow response over time. The recorded data were then subjected to two regression models (linear and polynomial) to obtain more information about the system. Next, the flow and voltage data over time were submitted to the PROCEST function in Matlab, which estimated a transfer function for the process using the input data. In this tool, various models for the system were calculated and compared based on their error values and fit to the data. Finally, the most suitable transfer function (a first-order function with a zero to the left of the real axis) was inserted into both an open-loop and a closed-loop system (both simulated using the SIMULINK simulation tool) and subjected to four classical tuning methods (ZieglerNichols, Cohen-Coon, and Trial and Error), whose performances were compared by observing their respective responses to a step disturbance inserted at the system input at the beginning of the simulation. Results showed similarly positive outcomes for Ziegler-Nichols tuning in openloop and Trial and Error tuning, both with quantitatively irrelevant steady-state error and rapid rise and settling times. The remaining methods were identified as inadequate for the identified system. Overall, the implementation was successful, and the system adapted well to both PID and PI control.