📊 Dashboard#
The STAMM Dashboard is an interactive web application developed with Plotly Dash, designed to serve both process control engineers and machine learning specialists working with soft sensors in industrial production environments. It provides a unified interface to visualize real-time process data, manage model deployments, and evaluate performance under operational conditions.
The dashboard is organized into four main sections, each focusing on a specific aspect of soft sensor supervision and analysis:
🗃️ Data Sources: this section displays comprehensive information about all experiments both ongoing and completed along with statistics describing the generated process data. Each experiment is linked to a specific project, enabling structured tracking across multiple workflows. If desired, users can connect to the IBISBA Hub, enriching the view with additional project-level details and metadata about the associated experiments.
🤖 Soft Sensors: in this section, users can select and deploy any soft sensor available in the Model Registry to operate in real time alongside a running experiment. Real-time plots display the simulation results of the selected soft sensor together with process variables such as sensors and actuators. Metadata about each deployed soft sensor — including configuration, version, and status — are also available for inspection.
📈 Monitoring: this section provides continuous monitoring of deployed soft sensors using both univariate and multivariate drift detectors. These detectors are part of a standalone Python package that provides interpretability through metadata and drift scores. Additionally, train–test density comparison plots allow users to visually compare the distribution of input variables in current operations versus those used during model training, supporting early detection of data shifts or process anomalies.
🧩 Simulation Assessment: once an experiment has finished (in offline mode), this section allows users to label and analyze sensor, actuator, and soft sensor simulation data. This functionality helps investigate abnormal behaviors, such as sensor faults or simulation errors, by enabling detailed post-run evaluation and traceability of system performance.
Explore the source code and contribute on GitLab:
This section provides a user-friendly guide for each part of the STAMM Dashboard. For every section, you will find a clear explanation of its main features. Also provides a collection of short video tutorials designed to help users explore and operate the STAMM Dashboard efficiently.
Each video focuses on a specific section of the platform, demonstrating its main functionalities through practical examples.