Air Pressure System is a vital component of any heavy-duty vehicle. It generates pressurized air that is used for different tasks such as braking, gear changing, etc. making it a very important subject of maintenance. Air Pressure System failure is common in heavy vehicles and the service and maintenance costs for such failures are high. We monitor the health of this system using sensors.
These sensors provide the company with real-time data. As these machines usually work in harsh environments, the sensors sometimes return abnormal data, which confuses the engineers.
To save cost and labour the company wants engineers to be sure about condition of air pressure system. So now we have a binary classification problem in which the affirmative class indicates that the anomaly was caused by a certain component of the APS, while the negative class indicates that the anomaly was caused by something else. If the anomaly was caused by APS component then engineers will repair or replace it.
- Building a robust machine learning training pipeline.
- When new training data becomes available, a workflow that includes data validation, preprocessing, model training, analysis, and deployment will be triggered.
- The final mode will be pushed into the S3 bucket. When the new model is trained the EC2 instance will compare the default model(which is already in S3) with newly trained validation data. If the new model performs better than the previous one, it will be pushed to S3 and the previous one will be deleted.
Step 1
git clone https://github.com/BackRoadAI/Detecting-IOT-Sensor-Fault-using-Machine-Learning.git
Step 2
conda create -p sensors python==3.7.6 -y
Step 3
conda activate sensors/
Step 4
python setup.py install
