A single line added to a PLC program, a temperature value changed in a recipe or a mandatory check step switched off in MES touches product quality far more directly than a server update in the office. Yet at most sites, these changes don’t fully enter either the IT change process or quality’s process change process. A maintenance technician intervenes at night, the line runs, no record is opened. Below, step by step, we build a process that works in the field and ties MES, PLC, machine recipe and test software changes to IATF 16949’s process change and product approval requirements.
The scope of change in production systems
The first job is to put in writing what counts as a “change”. Without a written definition, everyone uses their own. Each of the following objects can affect a product characteristic or process control:
- PLC and robot programs: logic, timer and counter values, safety and interlock conditions, robot paths.
- Machine recipes and parameter sets: process parameters such as temperature, pressure, time, torque and speed, and their permitted ranges.
- HMI/SCADA configuration: alarm limits, settings open to operators, permission levels.
- MES configuration: workflow steps, mandatory checks, product–recipe mapping, label templates, quality lock rules.
- Test and inspection software: limit values, test sequence, pass/fail logic.
- Infrastructure: operating system patches on HMI or station PCs, network changes, server migrations.
Infrastructure changes rarely affect product characteristics directly, but they can lead to line stoppages or data loss. A VLAN change on the cell switch breaks the connection between the PLC and MES, and the line stops with a “MES not responding” alarm. That’s why infrastructure stays in scope. Only its class is different.
The link between IATF process change and the IT change process
Clause 8.5.6.1 of IATF 16949 (control of changes – supplemental) requires a documented process for controlling changes that affect product realization. This process includes assessing the effect of the change, defining verification and validation activities, validating the change before implementation, evidencing the risk analysis and keeping verification/validation records. If the customer requires it, the change is notified to the customer and approval is obtained. Clause 8.5.6.1.1 separately covers temporary changes to process controls (for example, an error-proofing device being out of service and switching to an alternative control method).
On the IT side, there’s usually an ITIL-based change process: request, impact analysis, change advisory board, implementation, review. From the perspective of IT and information security compliance, the real problem is that the two processes run unaware of each other. The same change is opened in two places, or in none. The solution is a single entry point:
- Every change to production systems is opened with a single record (in the IT change tool or the quality system, but only one of them).
- “Is there a product/process impact?” is a required field in the record and is answered by process engineering.
- If there is an impact, the record is linked to quality’s process change flow; the need to update the PFMEA and control plan is assessed.
- If there’s no impact, the IT change process is enough, but this decision and its justification stay on record.
How IT changes are evidenced in audits is a separate topic; for the structure of request, approval and test records, see the article change management audit evidence. Here we focus on the flow specific to production systems.
Change classification decision tree
If you treat every change with the same weight, the process gets clogged. A week later, the maintenance team and operators start bypassing it; and you can’t really blame them. Classification determines the depth of approval and verification needed. The decision tree below is a starting example:
- Does the change affect a product characteristic, a special characteristic or a control in the control plan? If yes, Class A: process change, PFMEA/control plan review, trial production and assessment of customer notification.
- If not, does it move a process parameter outside the approved range or add a new parameter? If yes, it’s raised to Class A.
- If not, is it an adjustment within the approved range? If yes, Class B: approval by an authorized person and a record are enough; the recipe change history forms this record.
- If not, is it a software or infrastructure change with no product impact? (For example, HMI screen layout, an operating system patch.) If yes, Class C: IT change process, function test and rollback plan. If no question can be answered with a clear “yes”, the change is treated as Class A and process engineering makes the classification decision.
An emergency change isn’t a separate class; it’s the same tree run on a different timeline. In a breakdown response, the change is applied first, a record is opened by the next working day at the latest, and it’s classified and approved retrospectively using the questions above. If it turns out to be Class A, parts produced during the response are assessed separately. To see whether the emergency route is being abused, track the monthly number of emergency changes and records with overdue retrospective approval. If the number of emergency changes rises every month, the problem isn’t in the process, it’s in planning.
Class B rests on the concept of the “approved range”. The control plan and process parameter sheet must define the lower and upper limits for each parameter. If these limits are enforced technically in MES or the HMI, the operator can’t go outside the range without entering the change process. If the limit is only written on paper, it doesn’t exist on the shop floor.
PLC program and recipe version archive
The weakest link in the change process is often the version archive. Nobody knows whether “the program on the machine” and “the latest backup on the server” are the same. During a breakdown, the wrong version is loaded, and the line stops again on the next shift with the old fault. The table below compares different approaches:
| Approach | Advantage | Limitation |
|---|---|---|
| Folder on a file server | Simple, no extra cost | No comparison with the program on the shop floor; requires discipline |
| Git or a similar version control system | Seeing differences, history, signed records | Differences can’t be read in binary PLC project files |
| Industrial version management software | Automatic backup from the device and comparison | License cost, device compatibility check |
| Recipe management within MES | Product-linked recipes, approval flow | Covers only the parameters MES manages |
Whatever the tool, these principles don’t change: a single “approved version” is defined for each machine and controller; the approved version is associated with the change record number; the program on the shop floor is periodically read from the device and compared with the approved version. Engineering software from manufacturers such as Siemens and Rockwell offers project comparison features. On a line made up of devices from several manufacturers, vendor-neutral tools that take automatic backups from devices and produce difference reports may be preferred.
Git works well for text-based recipe or parameter files. An example of a simple workflow:
# Create a branch with the change record number
git switch -c CR-2025-0142-press3-temperature
# Edit the recipe file and review the diff
git diff recipes/PRESS3/product-4711.csv
# Commit with a message that includes the change record number
git commit -am "CR-2025-0142: PRESS3 4711 mold temperature 185->190 C (approved range 180-195)"
# After approval, merge into the main branch and tag it
git switch main
git merge --no-ff CR-2025-0142-press3-temperature
git tag -a PRESS3-4711-r12 -m "Commissioning approval: CR-2025-0142"
The change record number in the commit message is the only thread linking Git history to the change record in an audit. Don’t leave it out. A similar approach for configuration backup and difference tracking of network devices is explained in the article configuration backup with Oxidized; the logic applies to switches in the production cell too.
Verification and commissioning approval
8.5.6.1 requires the change to be validated before implementation. In production systems, this is a commissioning check whose depth depends on the class:
- Class A: trial production, first-part approval, measurement of the relevant characteristics, a process capability study if needed; verification of error-proofing devices.
- Class B: first-part check and confirmation that the recipe value was loaded correctly in MES/HMI.
- Class C: function test; software validation if measurement or test software is affected.
A rollback plan is written for every class: under which tag is the previous version in the archive, who does the restore, and how long does it take? If there are no answers to these questions, the change doesn’t start. The article validating measurement and test software covers the revalidation requirement for test and measurement software in detail.
Changes that require customer notification
Ultimately, which changes are notified to the customer is determined by the customer’s specific requirements (CSRs) and PPAP requirements. The PPAP manual published by AIAG lists, as examples of situations requiring notification, use of a different material or construction, production with new or modified tooling, moving tooling to another plant, a change of sub-supplier, restarting production with tooling that has been inactive for a long time and use of a new technique in inspection/test methods. For the current and complete list, go by the PPAP edition and CSR your own customer applies.
The practical consequences on the IT and automation side are:
- A change of test method in test stand or inspection software may require notification even if the acceptance criteria stay the same.
- Moving a line’s PLC program to the same type of machine at another plant may be assessed as a process relocation.
- Removing or disabling an error-proofing control through software, even temporarily, is managed under 8.5.6.1.1.
The last item causes the most problems in the field. A sensor fails, the maintenance technician bridges the bit in the PLC, and the line keeps running. The bridge stays for weeks. The effect of CSRs on IT processes is a broader topic; the article customer-specific requirements and IT explains how to derive IT actions from CSRs.
Sample change record
The fields below are an example of the minimum content of a record kept for a recipe change:
| Field | Example content |
|---|---|
| Record no. | CR-2025-0142 |
| Object | PRESS3, recipe for product 4711 |
| Change | Mold temperature 185 → 190 °C |
| Reason | Improvement linked to a surface defect corrective action record |
| Class | B (within the approved range of 180–195 °C) |
| Impact assessment | PFMEA and control plan unchanged; process engineer approval |
| Verification | Surface check of the first 5 parts, measurement record attached |
| Version | Previous r11, new r12; archive tag |
| Rollback | r11 is loaded; owner: maintenance shift supervisor |
| Customer notification | Not required; justification with CSR clause reference |
This record must be able to be matched with the program on a machine the auditor picks at random on the shop floor. The auditor reads the recipe revision on the HMI, and you show the tag in the archive and the change record. If they don’t match, that’s where the finding is. To adapt change and configuration processes to the production environment, take a look at our change, configuration and patch management and IT/OT compliance at manufacturing sites services.
Frequently asked questions
Does a parameter adjustment the operator makes from the HMI go through the change process?
An adjustment within the approved range doesn’t need a separate change record; but who made the adjustment, when, and from which value to which value must be recorded in the system. Going outside the approved range should be technically blocked and redirected to the process change process.
How is the change process run during an emergency breakdown response?
The change is applied first, a record is opened by the next working day at the latest and it’s classified retrospectively. If the change affects a product characteristic, parts produced during the response are assessed separately.
Is managing PLC programs with Git enough?
It’s useful for archiving and history, but as most PLC project files are in binary format, you can’t read the differences in Git. For difference analysis, you need the compare function in the manufacturer’s engineering software or industrial version management tools. Periodically comparing the program on the shop floor with the archive is a separate control.
Should every software change be notified to the customer?
No. The notification requirement is determined by the customer’s specific requirements and the PPAP rules. Changes that alter a product characteristic, process method or test technique usually require notification; infrastructure changes with no product impact usually don’t. In every case, the decision and its justification are kept on record.