Predicting Asset Maintenance Failure Using Supervised Machine Learning Techniques

Abstract

Maintenance activities can be broadly divided into three major categories and are corrective, preventive, and predictive maintenance. Our research focused on condition monitoring which is a form of predictive maintenance for brake pad failure for heavy-duty vehicles asset. The failure of a machine can stop production and cause a huge number of losses of money and people, moreover, it may take several months to order a new one. At the same time, excessive maintenance actions may slow production. Existing works of literature on predicting maintenance were studied in this research. Different machine learning techniques have been used for predicting maintenance, and to the best of our knowledge, Neural Network was only used for the prediction of the brake pad failure. Neural Network makes accurate prediction if the dataset is very large and also consume a lot of computational power. However, due to the fact that the problem is a classification problem, it is a necessity to carry out performance check of the best supervised model for the dataset downloaded from GitHub. Gaussian Naïve bayes, Decision tree and K-Nearest Neighbour are used to check for accuracy of our dataset. The dataset was divided into training and testing data where the training data has larger rows than the testing data. We then compared the performances of the selected supervised algorithm. Python is the preferred language used in this research. For our result, we showed that Decision tree performed well more than Gaussian Naïve bayes and K-Nearest Neighbour with an accuracy of 95%.