Series-built machines often arrive at the factory with the same IP plan: every filling machine’s PLC is 192.168.0.1 and its HMI 192.168.0.2. With a single machine there is no problem. When a second machine is connected to the same network, a conflict is inevitable; the symptom is usually one machine’s values appearing on another’s screen in SCADA, or the connection dropping and coming back at random. Below, we describe the 1:1 NAT, masquerade and routed integration options for connecting machine builder (OEM) networks to the factory network, NAT’s limits with industrial protocols, and the technical items that must be put in writing with the OEM at the ordering stage.
Why IP conflicts are inevitable with series-built machines
From the OEM’s point of view, the same IP plan is a deliberate engineering choice: the PLC program, HMI project, drive configurations and test procedure stay the same on every machine. The factory acceptance test (FAT) and site acceptance test (SAT) are carried out against this fixed plan; the remote support team knows every machine by the same addresses. When an IP change is requested, the OEM sees it as additional engineering work and a risk in terms of warranty and support responsibility. This attitude is understandable.
On the factory side, however, each machine must be reachable with a unique address from SCADA, MES, the historian and maintenance stations. What’s more, the range used inside the machine (often 192.168.0.0/24 or 192.168.1.0/24) may also be in use on the factory’s office network or in another cell; in that case, the problem arises not only between machines but also between the office and the machine. Setting up the factory’s own address plan is the subject of the article IP addressing and device naming standards for factory networks; the issue here is connecting machine networks that come from outside that plan without conflicts.
Integration options: 1:1 NAT, masquerade and routing
A machine router (usually an industrial router or security appliance with NAT and firewall features) is placed between the inside of each machine and the factory network. There are three basic methods:
- 1:1 NAT (static NAT, subnet mapping): The machine’s internal subnet is mapped one-to-one to a unique subnet in the factory plan. 192.168.0.10 appears on the factory side as 10.30.11.10; the host part is preserved. Inbound and outbound connections both work.
- Masquerade (NAPT): All traffic leaving the machine is translated to the router’s single factory IP. Connections from outside into the machine can only be made with port forwarding; if there are several PLCs serving on the same port (such as TCP 102 on two PLCs), a different external port is needed for each, and most SCADA drivers cannot handle this.
- Routed integration: The machine is readdressed with a unique subnet from the factory plan; the router only does routing and firewall work. There is no NAT, protocol problems are minimised, and real addresses appear in the logs.
| Criterion | 1:1 NAT | Masquerade | Routed |
|---|---|---|---|
| IP change on the machine | None | None | Required |
| Access to the machine from outside | To all devices | Via port forwarding, limited | To all devices |
| Factory IP consumption | One subnet per machine | One IP per machine | One subnet per machine |
| Protocol compatibility | Good for most TCP protocols | Good only for outbound data flows | Best |
| Typical use | Fleet of series-built machines | Machine that only sends data | New projects, OEM-compatible |
1:1 NAT is a common choice for a fleet of series-built machines. Masquerade is enough when the machine only sends data outwards (e.g. publishing to a broker via MQTT). Routed integration is the cleanest structure in the long term and can be made mandatory through the specification in new purchases. Trying to use masquerade for “access from the factory to the machine” turns the port forwarding table into a pile nobody can untangle.
The limits of NAT with industrial protocols
NAT only changes IP and TCP/UDP headers. It does not fix addresses carried inside the protocol, or Layer 2 traffic. The ones that cause problems in the field:
- PROFINET RT and DCP: These are Layer 2 protocols and do not pass through a router. A controller on the factory side cannot drive the I/O devices of a machine behind NAT via PROFINET; if I/O between machines is needed, a PN/PN coupler that connects to the two networks separately as an IO device without joining them, or TCP-based data exchange, is used. Device discovery by engineering software does not cross NAT either; the connection is established by entering the mapped IP address.
- EtherNet/IP: Explicit messaging (TCP 44818) usually works over 1:1 NAT. Implicit I/O (UDP 2222) uses multicast by default and does not pass through a router; if I/O between machines is needed, go for unicast connections and devices whose NAT support is explicitly stated. Because the ListIdentity response carries the device’s own IP, discovery tools may show the internal address. Do not trust the address shown on screen.
- OPC UA: The server returns its own internal hostname or IP address in the endpoint list. If the client tries to connect to this address, it fails behind NAT; the typical symptom is that discovery succeeds and the connection drops right after. The solution is to replace the endpoint URL on the client with the discovery address, or to define an alternative hostname on the server; the address fields in the certificate are adjusted accordingly. The certificate side is covered in the article OPC UA security settings.
- OPC Classic (DCOM) and active FTP: They use dynamic ports and carry addresses in the payload; they are problematic with NAT. Rather than forcing them through NAT, it is healthier to put an OPC UA or MQTT gateway on the machine side.
- Modbus TCP and S7comm: They carry no IP address in the payload; these are the protocols that come closest to working without problems over 1:1 NAT.
The machine router and the position of the security boundary
The machine router does not just translate addresses; it is the boundary of the machine zone. That is why three decisions are made right at the start:
- Ownership: Does the OEM or the factory provide the router? For a device supplied by the OEM, who holds the management password, firmware updates and authority to change rules must be written down. The preferred model is for the device to be chosen from the factory standard and managed by the factory.
- Rule policy: The router should have not only NAT, but a default-deny policy and allow rules written according to the communication matrix. The router also acts as the cell firewall; for placement and mode selection, see the article cell-level firewalls.
- Remote access path: An OEM putting its own cellular modem or remote access box inside the machine opens a channel that bypasses all of the factory’s segmentation. You cannot know this without opening the cabinet door and looking. Remote support should be provided over the factory’s approved access path; the security operations side is covered in the guide machine builder remote access.
Address mapping table and documentation
With 1:1 NAT, how the factory side reaches the mapped subnet is designed separately. There are two ways: defining the mapped subnet as a static route to the router’s outside interface on the factory’s Layer 3 switch, or choosing the mapped addresses from the VLAN where the router’s outside interface sits and using proxy ARP on the router. A static route is more readable and does not enlarge the broadcast domain. We prefer the static route.
Example mapping table (addresses are examples):
| Machine | Internal subnet | Mapped subnet | Router outside IP |
|---|---|---|---|
| Filler-01 | 192.168.0.0/24 | 10.30.11.0/24 | 10.30.1.11 |
| Filler-02 | 192.168.0.0/24 | 10.30.12.0/24 | 10.30.1.12 |
| Filler-03 | 192.168.0.0/24 | 10.30.13.0/24 | 10.30.1.13 |
On a Linux-based machine router, this mapping is written with the NETMAP target, and the route on the factory Layer 3 switch in Cisco IOS syntax, as follows (eth0 factory side, eth1 machine side):
# Machine router: Filler-01
iptables -t nat -A PREROUTING -i eth0 -d 10.30.11.0/24 -j NETMAP --to 192.168.0.0/24
iptables -t nat -A POSTROUTING -o eth0 -s 192.168.0.0/24 -j NETMAP --to 10.30.11.0/24
# Factory L3 switch
ip route 10.30.11.0 255.255.255.0 10.30.1.11
ip route 10.30.12.0 255.255.255.0 10.30.1.12
ip route 10.30.13.0 255.255.255.0 10.30.1.13
On commercial industrial routers, the same structure is set up via a “1:1 NAT” or “subnet NAT” menu. Whatever the method, the mapping table is entered at the same time into the IPAM record, the SCADA driver configuration and the machine documentation. If the mapped address lives only in the router configuration, the team gets lost between internal and external addresses at the first fault diagnosis. A mapping that preserves the host part (192.168.0.10 ↔ 10.30.11.10) makes diagnosis noticeably easier. The reference sources are RFC 3022 for the basic definition of NAT behaviour and the netfilter project page for the NETMAP target.
Items to add to the OEM technical specification
Most NAT problems can be prevented at the ordering stage; once the machine has arrived on site, every change means extra cost and schedule slippage. Items to add to the purchase specification:
- The list of IP address, subnet mask and gateway for all devices inside the machine is included in the handover documentation; a gateway is defined on all devices.
- The machine must be readdressable to the subnet the factory specifies, or the OEM demonstrates operation behind 1:1 NAT during the FAT.
- A list of protocol, port and direction for all services exposed outside the machine (communication matrix) is provided; unused services are disabled.
- If real-time I/O is needed between machines or with the factory controller, the method (coupler, unicast implicit connection, etc.) is stated explicitly in the quote.
- On OPC UA servers, the endpoint hostname and certificate address fields can be configured according to the names the factory provides.
- There is no cellular modem, wireless access point or remote access device inside the machine without the factory’s knowledge; remote support is provided over the factory’s access path.
- Management authority over the machine router, responsibility for firmware updates and the configuration backup are among the handover items.
These items are carried over unchanged to the SAT checklist when the machine is accepted: the SCADA connection from the mapped address, historian collection, online access from the engineering workstation and the firewall’s default-deny behaviour are tested one by one on site. For connecting machine networks to the factory architecture, see our IT/OT integration consulting page, and for physical network design, our industrial Ethernet and factory networks page.
Frequently asked questions
Does putting machines with the same IP plan into different VLANs solve the conflict?
The machines are separated from each other, but SCADA or a historian on the factory side cannot route to the same IP address in two different VLANs. VLAN separation alone is not enough; NAT or readdressing is needed so that the machines appear to the factory with unique addresses.
If I use 1:1 NAT, can I connect to the machine PLC with TIA Portal?
Because device discovery is Layer 2, a PLC behind NAT does not appear in the accessible devices list. The connection is usually established by entering the mapped IP address manually; on the firewall, TCP 102 access is opened only to the authorised engineering workstation. Verify the manufacturer’s NAT notes for the PLC and software version you use.
Should the OEM or the factory provide the machine router?
Because the router is the security boundary of the machine zone, it is preferable for the factory to manage it. If the OEM supplies the device, handover of the management password, firmware updates, authority to change rules and the configuration backup must be written into the contract.
When is masquerade enough?
Masquerade is enough if the machine only sends data outwards, for example publishing to an MQTT broker or dropping files on a central server. If systems on the factory side need to connect to more than one device on the machine, 1:1 NAT or routed integration is chosen.