Predictive Maintenance in Delhi Metro

In the last place, I interned in (Delhi Metro), right before the corona lockdown started, I built a predictive analytics solution. Delhi metro has a humongous network of rails ( most dense in India ) with data being collected at a very fast rate. But most of it is unstructured and categorical data which is only readable by the maintainers in the workshops the train gets sent to on a monthly basis. So I built a model to clean the data, preprocess it and using 2 methods ( SPADE and ARIMA ) I found out the patterns in occurrences of faults in components (which component had a higher probability of failing after a failure in a specific component) and built a time series out of the fault data to build an ARIMA model out of it so that they can predict future faults! Although the accuracy of the model was relatively low because I was interpolating quite a few data points since we had fewer data to train my model (the data collection process of Delhi Metro is slow since the trains visit the workshop only once a month and I worked as an intern for only 2 months), once the complete data of the complete railway network gets collected, we will have a robust model to predict future component failures in trains.
We have also pushed the data to the cloud and created dashboards so that all stakeholders can easily access the data. The complete implementation was done by me in 2 months. :D