Edge Ingestion
We capture data at the edge of the operation, where it is generated, without depending on central connectivity. We deploy edge ingestion software on plant equipment that reads signals locally, holds them through network outages and forwards them once the connection is restored.
Industrial edge data ingestion, in short
Industrial edge data ingestion means capturing plant data where it is generated: next to the PLC, the drive or the sensor. An edge node (typically an industrial gateway on the OT network) reads the signals, applies a first layer of processing and delivers them to the company data layer without depending on central connectivity.
Processing at the edge solves three practical problems. Latency: local reactions happen on the plant network, not after a round trip to a remote server. Volume: filtering by change of value and aggregating by time windows means transmitting information, not noise. Resilience: a local store-and-forward buffer keeps capturing during connectivity outages and forwards the backlog in order once the link recovers, so the production record has no gaps.
A typical architecture combines an acquisition gateway speaking the native protocols of each source (OPC UA, Modbus, S7, Ethernet/IP, IO-Link, even local files or SCADA databases), a disk-based buffer, and a normalisation step that maps raw tags to a common model with readable names, units and plant hierarchy, usually published over MQTT. Direct-to-cloud transmission remains reasonable for low-frequency, cloud-native sources with stable connectivity; most plants combine both. Our scope is data connectivity: we connect the equipment you already have to a common data layer, without replacing PLCs or sensors. See PLC connectivity, MQTT integration and the Captia Connect overview.
How it connects to the system
Connect enables data flow to AI and Energy and supports execution that Service digitises in the business layer.
Key concepts
Other solutions
- PLC Connectivity
We connect PLCs from any manufacturer and protocol to the Captia unified data layer, making control-layer information reliably available to the rest of your systems, from process historians to dashboards, without altering the control programs.
- OT/IT Integration
We build the bridge between the operational world (OT) and the information systems world (IT): a data layer that reads plant information and delivers it to business systems, with an architecture designed not to interfere with production.
- MQTT Integration
We implement MQTT brokers and data flows to connect sensors and industrial devices at scale, with topic hierarchies designed around the plant and options such as Sparkplug B when several systems consume the same data.
Frequently asked questions
- What is industrial edge data ingestion?
- It is the capture and first processing of plant data at the edge of the operation, where it is generated: next to the PLC, the sensor or the line. Instead of sending everything raw to a central server or the cloud, an edge node reads, filters and normalises the signals before transmitting them, without depending on central connectivity.
- What happens to the data if the cloud connection drops?
- A well-designed edge ingestion architecture includes a local buffer: the edge node keeps reading the equipment and stores data on disk until the connection recovers, then forwards the backlog in order. Capture does not stop, because it does not depend on the cloud being available at every moment.
- Do I need to replace my PLCs or sensors for edge ingestion?
- Generally no. The edge gateway speaks the protocols your equipment already uses: OPC UA, Modbus, Profinet, S7, Ethernet/IP or sensor signals. Our work is to connect what already exists to a common data layer, not to replace equipment. Only very old hardware with no communication interface at all requires evaluating alternative reading methods.
- When is edge processing better than sending everything to the cloud?
- When the signal volume makes raw transmission expensive or slow, when the plant needs low-latency reactions, or when connectivity is unstable and gaps in the historical record are unacceptable. If sources are few, cloud-native and the link is reliable, direct transmission may be enough.