Investigating Strategies and Parameters to Predict Maintenance of an Elevator System

In this era of automation, our lives are surrounded by machines, be it a mobile phone or an elevator. We humans become careless when it comes to the maintenance of the machine. From the customer’s perspective, until an elevator is not working, nobody tends to care. This carelessness, in the long run, can result in loss of human life as well as financial losses. Elevators require maintenance and safety. To overcome both, the machine requires timely maintenance, and it can be executed with the precise product vision with the help of predictive maintenance. It not only predicts future failure but also pinpoints the issues in complex machinery and gives better results in terms of preventive maintenance. The conventional predictive maintenance machine learning techniques are established on feature engineering. It is the manual formation of precise features using domain proficiency and similar methodologies. Due to this, models are hard to reuse because feature engineering is specific to the problem structure and the data available, which can vary from one place to the other. Deep learning methodologies provide better results due to the extraction of new deep features from the dataset compared with the existing features. This work reviews the extant literature as well as showcases the implementation of random forest classifiers on the open-sourced dataset. In our model, an average accuracy of 91.50% was obtained. The dataset consisted of sensor data, which were recorded on the basis of maintenance actions being taken.