I am Senanur Cetin, the author and maintainer of the NEXUS Smart Factory project. I built this application and published it as an open-source repository at https://github.com/senanurcetin/smart-factory-app.
The project focuses on predictive-maintenance analytics, specifically designed for maintenance and reliability teams who need to make data-driven decisions about which assets to inspect first. By turning sensor telemetry into a ranked maintenance work queue, the application helps prioritize equipment that shows the highest risk of failure.
The Problem: Prioritizing Maintenance Tasks
In industrial environments, maintenance teams often face a flood of sensor data without a clear way to prioritize their actions. The challenge is to transform raw telemetry into a prioritized list. Reliability engineers need to know which assets are at risk right now.
The NEXUS Smart Factory project addresses this by using a trained failure-risk model to process sensor data and output a work queue where assets are ranked by their predicted risk level. This is intended for engineers who need to decide which assets to inspect first to prevent downtime.
The Approach: Predictive Analytics and Dashboards
The project utilizes two primary datasets to demonstrate different aspects of predictive maintenance:
- UCI AI4I 2020 dataset: Used for training a failure-risk model that identifies potential equipment breakdowns.
- NASA C-MAPSS dataset: Used for a separate remaining-useful-life (RUL) regression study.
Failure-Risk Model and Dashboard
The core of the application is a model that analyzes sensor telemetry to predict the likelihood of failure. This model powers a plant dashboard built with Flask. Instead of just showing raw numbers, the dashboard presents a ranked maintenance work queue. This allows a user to see exactly which machines require immediate attention based on the model's risk assessment.
RUL Regression Study
In addition to failure risk, the project includes a study on Remaining Useful Life (RUL) using the NASA C-MAPSS data. This regression study provides a different perspective on asset health by focusing on the remaining lifespan of the equipment.
Measured Results
Based on the project documentation at https://github.com/senanurcetin/smart-factory-app, the implementation achieves the following:
- Ranked Maintenance Work Queue: The system successfully turns sensor telemetry into a prioritized list for maintenance teams.
- Failure-Risk Prediction: A trained model identifies assets at risk of failure.
- RUL Regression: The project includes a study for predicting remaining useful life using the NASA C-MAPSS dataset.
Limitations and Constraints
While the project provides a functional end-to-end pipeline, there are certain limits to the current implementation:
- Dataset Specificity: The project is built specifically around the UCI AI4I 2020 and NASA C-MAPSS datasets.
- Target Audience: The tool is designed for maintenance and reliability teams and engineers reviewing the portfolio project, rather than as a general-purpose data visualization tool.
How to Run the Project
To explore the code and run the application, you can set it up to run it locally. The project is structured as a Flask application that serves the plant dashboard.
- Access the repository at https://github.com/senanurcetin/smart-factory-app.
- The application uses Python and Flask to serve the dashboard.
- The project includes the necessary logic to process the UCI AI4I 2020 and NASA C-MAPSS datasets to generate the maintenance queue and RUL study.
- Once the environment is ready, you can run it locally to view the ranked maintenance work queue and the predictive analytics interface.
