cvijit/Hydraulic_System

The key role in developing and successfully achieving high-quality results in business is by minimising the cost and maximising the profits. This is possible only if proper resources are optimized and implemented. Failure prognosis is a part of predictive maintenance where data science field is involved in predicting the conditions of a system. With proper machine learning techniques, the monitoring devices can easily replace traditional monitoring devices. In this research, proper fault diagnosis is carried out on the components and the stable conditions of a hydraulic system. A dashboard is created where all the data point is explored by showing their distribution, the correlation matrix, their importance in predicting the conditions and the outliers. Chi-square test of independence is calculated to define the relationship between the categorical values. The scaling and dimensionality reduction step was done by using Quantile Transform Scalar and UMAP technique respectively. The model building and evaluating stages were implemented in Python where RandomisedSearchCV is used for hyperparameter optimization in six classification algorithms. Results showed that using gradient boosting decision trees algorithm helped in achieving greater accuracy than any other machine learning models. The web app was deployed for the research project using Heroku and the dashboard created for exploratory data analysis was published to web using Shiny apps in R.