Edge and I/O

Field devices, edge controllers and machine vision. Increasingly where data acquisition happens without touching the control system at all.

What lives in this layer

This is ISA-95 Levels 0 and 1: the sensors, actuators and I/O that touch the process, plus a newer class of device that sits beside the control system rather than inside it. The edge layer covers several kinds of hardware.

• Field devices and sensors. Photoeyes, proximity sensors, flow and temperature transmitters, and smart devices that report diagnostics over IO-Link or an industrial Ethernet network. Banner Engineering is a common name here, including wireless sensor networks for points that were never worth running cable to.
• Remote I/O. Modules that bring signals onto the network close to the equipment. Some newer units, such as Opto 22’s groov RIO, include built in networking, security and MQTT publishing, so a handful of signals can be collected and sent to a broker without a PLC in between.
• Edge controllers. Devices such as Opto 22’s groov EPIC combine control, I/O, a local HMI and software for data handling on one platform.
• Industrial PCs and gateways. Hardened computers that run protocol conversion, MQTT clients, local databases or SCADA at the machine. The major automation vendors, Siemens and Rockwell Automation among them, also offer edge platforms that sit next to their controllers and run applications against controller data.
• Machine vision. Smart cameras and vision systems, such as Cognex, for inspection, code reading and verification, which generate quality data that often never leaves the station.

We cover the wider landscape in edge solutions for industrial automation.

The problems we see most

Data trapped at the machine. Many machines run on older PLCs that cannot be modified safely, or that belong to an OEM who will not support changes. The data exists, but getting it means touching validated or warrantied code.

Polling that misses events. Reading controller tags on a timer works for slow values but misses short faults and stamps data with the wrong time. Edge devices that timestamp at the source and publish on change avoid most of this.

Gateway sprawl. Each project adds its own small box, configured by whoever installed it. Five years later there are twenty, with different firmware, default passwords and no owner.

Vision results that stay on the camera. Inspection systems reject bad parts but rarely feed the reason codes back to anyone who could fix the cause upstream.

The decisions this layer forces

Collect alongside, not through, when the control system is off limits. For monitoring, adding a few sensors and a remote I/O module is often cheaper and lower risk than modifying a legacy PLC program. We recommend it when the goal is visibility. We do not recommend it for anything that needs to control or interlock the process. That belongs in the controller.

Treat edge devices as managed assets. Every gateway needs an owner, a naming standard, a firmware plan and a place on the network diagram. If that cannot be agreed, we would rather install fewer devices.

Pick the edge for the job, not the brochure. A single remote I/O module publishing MQTT and a full industrial PC running local applications are very different commitments. We size the device to the use case and the skills of the team that will support it.

Keep a local copy. Edge devices should buffer and serve data locally so that a broken internet or WAN link does not mean lost production records. See why local data matters.

How it connects to adjacent layers

Edge devices sit next to the controls layer and must not undermine it. They communicate over the OT network, which is where most of the security risk of a new device shows up. Their data usually lands in SCADA and historians or a broker that is part of your plant data architecture, and from there feeds MES and batch for counts, states and quality results. MQTT is the common thread; our MQTT guide explains it.

Where Joltek helps

We design edge data collection as part of IT/OT integration work, so devices fit an overall architecture instead of adding to the sprawl. A plant systems assessment inventories what is already installed. When the goal is downtime visibility on machines you cannot easily modify, edge collection is often the fastest route to OEE and downtime improvement. When the work moves to build and install, we manage it as systems integration or oversee it as owner’s engineer.