Direct Answer
Integration between paper machinery and a plant's ERP or MES is a specification decision made at purchase, not a project started after installation. The line must publish defined data points, through an interface the plant can maintain, with documented register maps, units and failure behaviour. Buyers should name the protocol, the data set, the update model and the expected network load in the contract, and should verify the interface during factory acceptance using a simulator or test mode. The common failure is not a technical one: it is a machine that produces excellent paper products and no usable data, leaving planners to key numbers in by hand. Decide the data model before the deposit, test the link before shipment, and phase the rollout so that production starts with a working line rather than a stalled network project.
Opening Hook
The new converting line was installed in six weeks and connected to the plant network in six months. Production had run for a quarter with clipboards at the machine, because the interface existed on the wiring diagram and nowhere else. Nobody had asked what the machine would publish, in which protocol, or who would map the tags. The supplier had built a sound machine; the buyer had bought one without a data model. The correction costs an integrator's time and weeks of manual reporting that need not have happened. At yoco-group, we ask buyers for their data list at quotation stage, because a converting line is bought twice: once as iron, and once as information.
What Integration Actually Means on a Converting Line
Integration is not a cable and not a dashboard. It is an agreement about which facts leave the machine and what the plant does with them.
| Layer | Question it answers | Typical consumer |
|---|---|---|
| Device and control | What is the machine doing now? | Control cabinet, PLC |
| Supervisory | What did this shift produce? | Line HMI, local reporting |
| MES | Which order, which batch, which quality result? | Production management |
| ERP | What did it cost, what do we owe, what do we ship? | Planning, finance, logistics |
| Analytics | Where is the recurring loss? | Continuous improvement team |
Most disappointment is a layer mismatch. A buyer expects ERP-level cost reporting and receives device-level signals, then discovers that nobody owns the translation between them. The remedy is to write the integration specification as a list of questions with named consumers, and to decide at purchase who maintains the mapping. A machine that reports state only to a local lamp is not a step down; it may be exactly right for a plant without a MES. The error is paying for capability nobody will operate.
Specifying the Data Set Before the Deposit
The data set is the most negotiable and most neglected part of a machinery contract, and it is cheap to fix at quotation stage.
| Data point | Typical granularity | Why planners want it |
|---|---|---|
| Machine state | Running, stopped, fault, changeover | Availability reporting |
| Output count | Per shift, per order, per hour | Performance measurement |
| Reject count | Counted or classified by defect | Quality and yield |
| Speed | Actual against setpoint | Drift detection |
| Order or batch ID | Operator entry or scan | Traceability and costing |
| Key process parameters | Temperature, pressure, tension | Quality correlation |
The discipline that keeps this list honest comes from value stream thinking: collect only what changes a decision. The Lean Enterprise Institute's production resources make the same argument about measurement generally, that a metric which nobody acts on consumes attention without returning value. Applied to a machinery purchase, it means the buyer should be able to name the person who will read each data point and the decision it will drive. Anything that fails that test can be deferred to a later phase, which keeps the first integration small enough to finish.
Once the data set is fixed, the interface options should be compared on maintainability rather than novelty. Our PLC and HMI control system guide sets out how controllers and local interfaces are typically structured on paper converting equipment, which is the layer most integration projects actually touch.
Data: ISO publishes industrial data and automation standards that define how control systems exchange information and how automation terminology and data models are structured across equipment from different vendors.
Judgment: Specify a standards-based interface and documented data model at purchase, because a proprietary link narrows the buyer's future choice of integrator and raises the cost of every later change.
Source: ISO — Industrial Data and Automation Standards (2024)
Interface Options and How to Choose Between Them
Buyers usually face three practical routes, and the right answer depends on the plant's skill base more than on the machine.
| Interface route | Strength | Watch out for |
|---|---|---|
| Fieldbus or industrial Ethernet to the PLC | Fast, deterministic, already in the cabinet | Needs plant networking knowledge |
| OPC-style server on the line controller | Standardised data access for higher systems | Licensing and naming conventions |
| Database or file exchange | Simple, tolerant of weak networks | Latency, duplication risk |
| Vendor cloud portal | Quick visibility, low local effort | Data ownership and recurring cost |
None of these is universally correct. A plant with an experienced controls engineer and a modern network can take the fieldbus route and gain speed and traceability. A plant without that skill base is often better served by a supervised file exchange that it can support alone, accepting slower updates. The questions to settle in writing are the same in every case: who owns the data, who maintains the mapping, what happens when the link drops, and whether the machine continues producing during a network outage. That last question is a safety-of-production issue, and the answer should always be that the line keeps running.
Verifying the Interface at Factory Acceptance
Integration defects found at the supplier's works cost an afternoon; the same defects found at commissioning cost a production week.
| Verification item | How to test it | Pass condition |
|---|---|---|
| Data points present | Connect to the test interface and enumerate tags | Every agreed point returns a value |
| Units and scaling | Compare against a manual reading | Values match within stated tolerance |
| State transitions | Start, stop and fault the line | State changes appear in order |
| Order or batch entry | Load a test order | Identifier flows to the higher system |
| Link loss behaviour | Disconnect the network cable | Line keeps running, buffer holds |
| Recovery | Reconnect after an outage | Data resumes without duplication |
The two tests buyers skip most often are link loss and recovery, and they are the two that decide whether the plant trusts the system. A line that stops when the network blinks will be disconnected by operators within a week. Run these checks as part of the acceptance protocol on the same day as the mechanical run, and record the results in the same signed report. Our automation guide for paper machinery describes where sensing, actuation and data collection typically sit across converting equipment, which helps a buyer decide which signals are worth the wiring.
Data: The U.S. Department of Energy's Advanced Manufacturing Office publishes resources on manufacturing systems and productivity, including how measurement and process control affect industrial energy and output performance.
Judgment: Justify each data point by the decision it supports, because instrumentation that produces no action adds maintenance and network load without improving output.
Source: U.S. Department of Energy, Advanced Manufacturing Office — Manufacturing Systems and Productivity Resources (2024)
A Phasing Plan That Does Not Stall the Plant
Rollouts fail when everything is scheduled at once and the plant has no working baseline.
| Phase | Scope | Typical duration logic |
|---|---|---|
| Phase 0 | Data set agreed, interface chosen, tags mapped | Completes before deposit release |
| Phase 1 | Machine to supervisory layer, local reporting | Tied to commissioning |
| Phase 2 | Supervisory to MES, order and batch flow | After stable production |
| Phase 3 | MES to ERP for costing and planning | After data quality is trusted |
| Phase 4 | Analytics and improvement loops | Ongoing |
Phasing is not procrastination; it is risk management. Each phase should end with something the plant uses, so that the next phase is funded by demonstrated value rather than by a project plan. Two governance rules support this. First, never let an integration phase block production start; the machine should run and report locally from day one. Second, assign a single owner for the data model across phases, because hand-offs between integrators are where mappings drift and duplicate identifiers appear. Buyers who follow both rules usually reach Phase 3 without the reporting backlog that makes integration projects look expensive.
Data: UL Solutions evaluates industrial control and equipment safety, documenting how control systems and their interfaces are assessed against defined safety requirements.
Judgment: Keep the integration boundary inside the equipment safety concept, because a network change that alters control behaviour should be reviewed as a safety-relevant modification rather than a routine IT task.
Source: UL Solutions — Industrial Control and Equipment Safety Evaluation (2025)
The Bottom Line
Decide the data set before the deposit, choose the interface the plant can maintain, verify the link at factory acceptance including link loss and recovery, and phase the rollout so production never waits for a network project. A converting line that cannot report is a line you will measure with a clipboard.