Development of an IoT-supported predictive maintenance system for agricultural irrigation pumps: A machine learning approach

Periodic maintenance of irrigation pumps used in rural areas is costly and time consuming. By obtaining real-time information, potential problems can be detected in advance. The aim of this study is to instantly monitor the performance of the pump installed for agricultural irrigation using the Internet of Things (IoT) and to apply regression models to the system data and to determine optimum parameter values for the models with genetic algorithm. First, the automation system was installed and the temperature, humidity, current, voltage and pump vibration data were transferred to the server named Thingspeak and the data were visualized in the application created with the Qt user interface creation toolkit. Faults occurring in the system were instantly notified to the user through this application. Predictive maintenance of the system consists of a two-stage process in the field of machine learning. In the first stage, linear regression, polynomial regression, random forest regression (RFR) and support vector regression (SVR) models were evaluated. In the second stage, the genetic algorithm was used in hybrid structure to determine the optimal parameters of the polynomial regression, SVR and RFR models. Improvement of model parameters for the irrigation pump has been achieved using genetic algorithm. As a result, evolutionary hybrid random forest regression model was found to be the most appropriate model for the dataset.