Air Pressure System (APS) Sensor Fault Detection
- 0 Collaborators
The APS Sensor, a critical element in heavy-duty vehicles, facilitates brake pad pressure by converting compressed air into piston force for deceleration. This study investigates the correlation between component failures and the APS. ...learn more
Project status: Published/In Market
oneAPI, Artificial Intelligence
Overview / Usage
The Air Pressure System (APS) is one of the important component of a heavy-duty vehicle that uses compressed air to force a piston to help for pressure to the brake pads. This is more sustainable and easy availability. This is a Binary Classification problem, in which the positive class indicates that the failure was caused by a certain component of the APS, while the negative class indicates that the failure was caused by something else.
Methodology / Approach
- The Data generated by the APS Sensors in fixed time intervals is send to Kafka (Confluent Kafka), which is a live data streaming platforms to store the large amount of data received from the sensor.
- For flexibility, the data present in the Kafka topic is written into a MongoDB Database whenever the model needs to be trained.
- Finally, a .csv file is created by extracting the data from the MongoDB.
- Performing different experiments with Imputer (KNN Imputer, Simple Imputer),Scalers(Min-Max Scaler, Robust Scaler) and check which gives the best result when model training.
- The Dataset is balanced using oversampling techniques like SMOTE and TOMEK to remove noise created in the process of oversampling.
- Training the Data using different model like RandomForest,Logistics Regression, K-Nearest Neighbour using the One-DAL library of **Intel's ONE API and XGBoost Classifier **of XGBoost library
- Choose the trained model with best accuracy and less cost function.
Technologies Used
One API
Confluent Kafka
Mongodb
XGBoost