<![CDATA[The milling machine is a crucial aspect of the sugarcane agroindustry production system; a disturbed milling machine will cause a decrease in production efficiency, sap quality degradation, and excessive energy consumption. An early fault anomaly detection system through machine learning is a solution to overcome the problems in sugarcane milling machines. The purpose of this research is to propose a system architecture design for early fault anomaly detection in sugarcane agroindustry milling machines and to evaluate the performance of various machine learning models on historical sensor data, identifying the most promising approach. This study proposes a novel anomaly detection framework for sugarcane milling machines to advance smart monitoring in agro-industrial systems. Using an empirical dataset of 7,673 sensor instances (temperature, vibration, pressure, and humidity), and applying several machine learning algorithms (logistic regression, decision tree, and random forest), the framework integrates multi-sensor data to improve fault prediction and reduce downtime. The results showed that the random forest had the best accuracy, at 98.13%, followed by the decision tree, at 97.87%, and logistic regression, at 89.70%. Feature contribution analysis reveals that the vibration signal is the most dominant contributing factor among other features. The results show that machine learning is a potential approach for predicting faults in sugarcane milling machines, which can help the sugarcane agriculture industry make informed decisions in the event of disturbances in these machines.]]>
