Predictive Maintenance of Hydraulic System using Machine Learning Algorithms

This project proposes the predictive maintenance and fault tolerance technique to increase the accuracy in Hydraulic system using Machine Learning. Hydraulic based systems are mainly used in industries, brake systems in vehicles, air crafts etc. In simple terms Hydraulic system functions and tasks are performed through the pressure to the fluid. The proposed project is used to maintain health condition of hydraulic system. Because the hydraulic pressure given to small amount of fluid will create huge amount of power to start an engine or start the gear pump, because of complex task and external environment some fault will occur. This project using Machine Learning (ML) algorithms which can find fault in attributes and increase fault tolerance. For that Machine learning Classification algorithms that are Logistic Regression, K Nearest Neighbor (KNN), Decision Tree, Random Forest, Naïve bayes are used to predict the internal faults such as Valve condition, Internal pump leakage, Hydraulic accumulator/bar, Stable flag. If there is any fault occurred, it will let us know before 15-30 minutes. These algorithms are implemented using software tool Spyder IDE which is installed in the Raspberry pi 3 Model B+ controller. By using these Algorithms, the parameters such as precision, accuracy and recall will be calculated. And for external fault prediction Flame sensor, temperature, Vibration sensor, Voltage sensor are used and Cooler (Motor) for Cooling purpose. Hence the Fault prediction of Hydraulic system from Logistic regression achieved accuracy of 93%.