Engineering teams may have test results yet still lack suitable data for training artificial intelligence models. Siemens presents Simcenter Testlab 2606 as a way to close that gap by connecting measurement, validation, labeling, storage, and delivery within one environment.
The article identifies seven requirements: quality, validity, traceability, completeness, quantity, accessibility, and connectivity. Poor measurements can teach models to reproduce sensor errors, while missing context makes results difficult to interpret or compare. Siemens SCADAS acquisition systems retain calibration, connectivity, and overload information alongside measurements. Testlab Process Designer automates checks on recorded signals and processed results.
Consistent annotation makes those results understandable. The Descriptive Model and Template Editor lets teams establish metadata structures and prepare templates before testing. Operators then add information specific to measurements. Engineers retain access to that context during analysis and reporting, helping them identify what was tested and compare results meaningfully. Metadata can also be corrected or expanded as work progresses.
The new Schedule Acquisition workbook brings expert procedures into Testlab Neo. Engineers define instrumentation, measurement tasks, operator instructions, annotation, processing, and reporting before a campaign begins. Software then guides operators through the prescribed sequence. Reusing settings and procedures makes campaigns more comparable and reduces dependence on individual experience. External applications and Python scripts can be included for specialized tasks.
Where physical measurements are insufficient, component-based Transfer Path Analysis and virtual prototype assembly can supplement datasets with simulated results. Centralized Testlab Data Management makes measurements searchable by attributes rather than filenames.
Workflow Automation also gains direct access to this database in version 2606. Teams can define searches, retrieve matching measurements, reprocess them, and export results to AI training platforms such as RapidMiner AI Studio.
For engineers, the emphasis is on preparing reusable data throughout testing, reducing the cleanup required before machine learning can support prediction and design decisions.
