Glossary · Sensors and measurement
TensorFlow (industrial applications)
Also known as: TensorFlow Industrial, TensorFlow Lite
German: TensorFlow Industrial
In industrial AI, TensorFlow is an open-source machine learning framework originally developed by Google that is used to train models on sensor, image and process data and to run them on servers, industrial PCs or edge devices. The list term TensorFlow Industrial refers to such industrial use, not to a separate product edition.
- Sensors
- Vendor product
- AI
In one sentence
TensorFlow is an open-source ML framework used in industry to train and run models on sensor, image and process data, from cloud to edge.
Example
A team trains an anomaly detection model on vibration data with TensorFlow and deploys a converted, lightweight version on an edge device next to the machine.
How it applies
- Engineering: TensorFlow is used for visual inspection, anomaly detection, soft sensors and predictive maintenance. Models are trained on historical data and then deployed for inference, often on edge hardware with reduced model formats.
- Data: Model quality depends on the quality and context of the training data, including units, timestamps, sensor calibration and labeled events.
- Safety and compliance: A model's output is a statistical estimate. Using it in a safety function or a regulated product requires separate assessment; the framework does not make an application safe or compliant.
- Documentation: Document model version, training data, intended use, limits and monitoring, for example in a Model card, and state in operating instructions what operators should do when the model's output looks implausible.
Keep in mind: Product names, editions, supported platforms and features of TensorFlow and related tools change. Check them against the vendor's current documentation before referencing them in technical documentation.