This project explores the application of data mining and predictive analytics to enhance predictive maintenance (PdAM) in automobile industries, aligning with the Industry 4.0 approach. It includes a Python-based implementation for predicting car engine failure using sensor data.
The research paper "Automobile Industries using Data Mining and Predictive Analytics: An Industry 4.0 Approach" [cite: 105, 106, 107, 108, 109] proposes a PdAM framework to help the automobile industry identify patterns using sensor data and predict equipment conditions, specifically wear and tear, before failure occurs[cite: 108]. This project puts those concepts into practice by developing a model to predict car engine failure.
- Industry 4.0 and PdAM: The project leverages the principles of Industry 4.0, which emphasizes smarter decision-making through advanced technologies for data collection and interpretation[cite: 105, 106]. PdAM is a key component, using sensor data to predict maintenance needs[cite: 108, 109].
- PdAM Framework: The paper outlines a five-phase PdAM framework:
- Data collection from various resources (sensors, IoT, data warehouses, etc.)[cite: 110].
- Data pre-processing (cleaning, transformation, reduction)[cite: 111].
- Algorithm selection (classification, regression, association)[cite: 112, 3].
- Predictive maintenance model development[cite: 112].
- Model training and testing for accurate results[cite: 112].
- Data Mining Algorithms: The framework utilizes various algorithms, including classification, regression, and association, to support decision-making in the automobile industry[cite: 107, 117].
- Benefits of Predictive Analytics: The research emphasizes that PdAM helps in early awareness of machine health, preventing damage, reducing downtime, improving safety, understanding failure causes, and increasing productivity and revenue[cite: 113, 114, 115, 116, 117, 118, 119, 125, 23, 24].
The Python implementation (PdAM.ipynb) demonstrates a simplified version of the PdAM framework, focusing on predicting engine failure from sensor data.
- Pandas
- NumPy
- Seaborn
- Scikit-learn (sklearn.model_selection, sklearn.linear_model)
- Data types are analyzed.
- Missing values are checked and handled.
- EDA is performed to understand relationships between sensor measurements and engine failure.
- The dataset is split into training and testing sets.
- A Linear Regression model is used to predict engine failure. But the Rsquare ans RMSE score was not good so used Classification Algorythm model as it was a binary dataset and that's why linear regression does not performed well. (Note: The research paper mentions various algorithms) [cite: 112, 3]
Car engine sensor data.csv: Dataset containing car engine sensor data.PdAM.ipynb: Jupyter Notebook with the Python implementation.
- Install the required libraries.
- Place the data file in the same directory as the notebook.
- Run the notebook to execute the analysis and modeling.
This project aligns with the research paper by:
- Applying the PdAM concept to a real-world problem (engine failure prediction)[cite: 108, 109].
- Using sensor data as a key input for predictive maintenance[cite: 110, 116].
- Implementing data pre-processing and modeling steps[cite: 111, 112].
- Demonstrating the use of machine learning (Linear Regression) for predictive analytics, although the paper suggests a broader range of algorithms[cite: 112, 3, 119].
- Focusing on the goal of predicting potential failures, which contributes to the benefits outlined in the paper, such as reducing downtime and improving safety[cite: 115, 23].
This project can be extended by:
- Implementing other algorithms (classification, etc.) as suggested in the research paper[cite: 112, 3].
- Incorporating more phases of the PdAM framework, such as data collection strategies and more advanced model evaluation[cite: 110, 112].
- Expanding the dataset with more variables to align with the inputs mentioned in the paper (temperature, sound, vibration, pressure, RPM, etc.)[cite: 24].