Industrial Data Collection and IIoT Integration
We determine the signals and sampling rates needed for production, quality, energy or maintenance. We record the unit, timestamp and quality of each collected value together with the data.
Edge bufferingTimestampsHistorianTime seriesNode-REDData quality
What problem does this service solve?
Collected data that cannot be trusted
Analysis based on data with no timestamp, a drifted clock or unclear units gives wrong results.
Gaps when the connection drops
Data is lost during network or server maintenance. Reports are left with gaps nobody can explain.
The “let's collect everything” approach
Collecting every signal at high frequency without a written purpose increases storage, network and PLC load; and which decision the data will be used for remains unclear.
Old machines that provide no data
Not even basic status and counter information can be obtained from a machine without a communication interface.
What we cover, and within what limits
Collecting production data means recording a machine’s status, counters, process values and alarms in a traceable way, at a period that serves a specific purpose (e.g. OEE, energy, quality, predictive maintenance). Reliable data has three requirements: synchronised time (NTP or a suitable time source), information showing the quality of the value, and a store-and-forward design that keeps data locally when communication drops. If there is a local buffer, its capacity, the outage length it can cover, which data is deleted when it fills up and how duplicate records are filtered out on resend are written down. Edge buffering is not built into every device; it is designed separately in the edge device, gateway or software used, and verified with a planned disconnection test. Data is stored in an industrial historian or a time-series database and served from there to dashboards, reporting or analytics tools.
Technical scope
Purpose and signal list
Which question are we trying to answer? Signals and sampling periods are defined according to the OEE, energy, quality or maintenance goal.
Method of collection
Collection options are compared: from the PLC via OPC UA/Modbus/S7, from the existing SCADA, or with additional sensors and I/O modules.
Time and quality
Time source and synchronisation, where the timestamp is assigned, value quality and unit information.
Edge buffering
Local buffer capacity, the outage length it can cover, its behaviour when full and the resend order. Lost data, late data and duplicate records are handled separately.
Storage layer
Choosing a historian or time-series database, compression/aggregation, retention period and backup.
Access and monitoring
Who will access the data, and with which tools; who gets an alert when the data flow stops.
Components of the data collection architecture
Collection (edge)
Reading from the PLC/machine, unit and name conversion, local buffer. Where appropriate, tools such as Node-RED can be used for flow orchestration.
Transport
Transmission to the central layer via OPC UA or MQTT; TLS and authentication.
Storage
Historian or time-series database; aggregation and retention policy.
Presentation
Dashboards, reports or analytics tools; access permissions.
Prerequisites: what we need to get started
We list missing information and access together in the first meeting and complete them during the survey.
- The decisions expected from the data and who owns them (e.g. maintenance manager, production manager)
- List of machines/PLCs and their communication capabilities
- A cabinet/room where the edge device or server can be located, and a network connection
- Time source information (NTP server, etc.)
- A decision-maker for retention period and data ownership
Deliverables
Use case and success criteria
Signal list (name, unit, period, source)
Collection and buffering architecture
Pilot machine integration and disconnection test results
Retention and backup policy
Data flow monitoring and operating documentation
Representative scenario: measuring downtime reasons with signals
On a hypothetical press line, downtime reasons rely on operator notes. Machine status, counter and alarm signals are collected at the edge layer with one-second resolution and synchronised timestamps. If the connection drops, data is kept locally for as long as the local buffer capacity allows. Measured downtime is compared with operator records over a defined period. Where they do not match, the signal definition is checked first.
This is a representative project flow; it is not a completed customer reference.
Service summary
OPC UA, MQTT and Data Integration
We read the data model a machine exposes via OPC UA and, in projects that need MQTT, distribute this data to applications through a broker.
Scope limit
Developing predictive maintenance or AI models is not part of this service; we build the reliable data infrastructure those efforts need. If sensors are to be added to a machine, installation is planned together with the machine manufacturer’s conditions and occupational safety rules.
How do we proceed?
No step that affects production is carried out until the maintenance window, configuration backup and rollback plan are written down.
Purpose
The use case and success criteria are written down.
Signal design
The signal list, periods and quality rules are defined.
Pilot
Collection is set up on a single machine, from reading to storage, and disconnection is tested.
Scaling
It is extended to other machines using a template.
Operation
Data flow monitoring and maintenance procedures are put into operation.
Questions about IIoT and Production Data Collection
Is data lost if communication drops?
It depends on the PLC, edge device and application used. Suitable storage and a store-and-forward design reduce the risk of loss for outages up to the length the local buffer can hold. Do not assume every device buffers on its own; verify it with a planned disconnection test.
How often should we collect data?
It depends on the purpose. For states and counters, event-based or per-second periods are usually enough; high-frequency measurements such as vibration need separate hardware and design.
Can data be collected from old machines without communication interfaces?
In most cases, yes. The machine’s existing signals (running, fault, part produced, etc.) can be read with a suitable I/O module or sensor. Any work that touches the control circuit is done with the approval of the person responsible for the machine.
Historian or time-series database?
Industrial historian products offer compression, OT integration and ready-made tools; time-series databases offer flexibility and integration with open tools. Existing systems, licensing and maintenance capacity are weighed together in the choice.
Work that can be planned together
OPC UA, MQTT and Data Integration
We read the data model a machine exposes via OPC UA and, in projects that need MQTT, distribute this data to applications through a broker.
Production Monitoring and Dashboards
We show line status on operator screens, recurring stoppages on maintenance screens, and shift and OEE indicators on management screens.
Edge Gateways and Protocol Conversion
To collect data from machines that use different protocols, we review the existing interfaces, manufacturer support and open data maps.
Knowledge Center
Technical guides for this service
Let's plan a technical survey for IIoT and production data collection
Briefly tell us which data you want from which machine and why, and which PLCs you use; we will respond with a written proposal covering scope, assumptions and deliverables.