Home

1. Predictive Maintenance (ML-based)

A key feature of the application is that it helps you create the labeled data. You provide two types of raw data:

1. Sensor Trend Data: This is your unlabeled feature data (temperature, vibration, etc.).
2. Event Logs: This contains records of historical failures and maintenance.

The application uses the event logs to automatically generate a failure_label column for your sensor data. It marks the data points occurring within a specified “lookback window” before a known failure as a “pre-failure” state (label = 1) and all other data as “normal” (label = 0).

See in action here:

2. Anomaly Detection

Most widely used technique to identify abnormal patterns in the incoming data; vibration, pressure, current, temperature and other process parameters.

See in action here:

3. Advanced Signal Processing and EDA

A comprehensive Streamlit web application for exploratory data analysis and signal processing that enables users to analyze time series data and waveforms without coding.

Features include CSV upload functionality, pre-loaded example datasets, and multiple analysis methods including FFT, STFT, Wavelet Transform, EMD, PCA, and correlation analysis with interactive Plotly visualizations. Perfect for vibration/current signal processing and IoT sensor data, offering both beginners and professionals a powerful tool for rapid data exploration and pattern recognition through an intuitive web interface.

Check out the below analyzer:

4. Feature Extraction, Visualization and Export

In machine learning, your model is only as good as its features. For time series data, this is critical. A model can’t easily distinguish a healthy machine bearing from a failing one using raw vibration data, but it can use engineered features like Kurtosis or Crest Factor to detect the tell-tale spikes that signal a fault.

Manually calculating these features is a major bottleneck in any data science workflow. The Time Series Feature Explorer is a web-based tool that automates this process. It’s a no-code solution designed to transform your raw signals into a structured, feature-rich dataset ready for analysis and machine learning.

Key Capabilities:

• Automated Feature Calculation: Instantly compute dozens of time-domain, frequency-domain, and trend-based features.
• Multi-Signal Comparison: Upload and compare feature sets from multiple signals side-by-side.
• Interactive Visualization: Analyze results with interactive plots, including grouped feature charts and FFT spectrums.
• Dimensionality Reduction: Perform PCA on your features and visualize the results to see how your data clusters.
• Data Export: Download the complete feature table as a clean CSV or JSON file.

5. Find patterns in time series

This app uses the Matrix Profile algorithm to automatically discover recurring patterns (motifs) and detect anomalies (discords) in time series data. You can either generate synthetic data (like ECG signals or sine waves) or upload your own CSV/JSON files, then the app computes a “matrix profile” that efficiently compares every subsequence in your data to find the most similar patterns and most unusual outliers. It’s perfect for any scenario where you need to understand hidden patterns and anomalies in time-ordered data without manual inspection.