Tool Summary: Multi-Input Multi-Output AI Trainer
✅ Problem statement
In manufacturing lines, rotating machines such as motors, pumps, and conveyors rely on bearings that gradually wear down. As they degrade, temperature and vibration start to rise — but detecting that trend early is often difficult without dedicated predictive systems.
Typically, engineers either inspect the data manually or build one-off scripts to fit a regression model — a process that’s slow, error-prone, and hard to reuse across machines.
With this demo, we set out to solve that pain point by developing a browser-based tool that lets users train an AI model to predict bearing temperature directly from operational parameters such as rotational speed, load, and vibration levels.
This makes it easy to explore relationships between process inputs and machine health indicators — without needing to code or configure an ML environment.
The AI Trainer automates the entire workflow: scaling, training, validation, and visualization, giving engineers immediate feedback on how well their data can support predictive maintenance decisions.
📌 Methodology
We used a shallow neural regressor implemented with scikit-learn’s MLPRegressor, designed to capture nonlinear dependencies between machine inputs and target variables such as bearing temperature.
Each dataset goes through:
- Feature Scaling with
StandardScalerto normalize input ranges. - Train/Test Split (80/20) for unbiased validation.
- Model Training on a neural network with two hidden layers (32 and 16 neurons, ReLU activation).
- Performance Evaluation via R² metrics and prediction–reality visualizations.
This simple yet effective setup allows quick prototyping of predictive models suitable for maintenance and process optimization.
🔍 Key Features
- 🧩 Multi-Input, Multi-Output Learning: Model several variables at once (e.g. vibration + temperature).
- ⚙️ Automated Preprocessing: Scaling and splitting handled automatically.
- 📊 Instant Visual Feedback: R² comparison and scatter plots of predicted vs. actual values.
- 🎯 Interactive Predictions: Enter new sensor values to forecast expected outcomes.
- 💡 No-Code Setup: Train and test models entirely from a browser session.
Sample data
Disclaimer: This tool is provided ‘as is’ without any warranties, and is intended for illustrative purposes only. Please verify all calculations independently before using them to make decisions.