8 mins

Legacy system modernization strategies for factory automation and machine controls

8 mins

Legacy system modernization strategies for factory automation and machine controls

Modernizing legacy systems in manufacturing

Legacy system modernization projects fail most often at the boundary between new application code and the deterministic machine controls that have run the line for decades.

This piece breaks down where that boundary is engineered correctly, and what to check for in a modernization partner.

Table of contents

What does it cost to keep running on a legacy system?

A 2026 survey by TXOne Networks covered about 550 senior decision-makers in 6 European countries. It found 3 things:

  • 60% say legacy Windows makes up at least half of their OT environment.

  • 51% had a security incident involving legacy Windows in the past 12 months. The prior year was 43%.

  • The top barrier to replacement is compatibility with legacy equipment (54%). Cost (38%) and downtime risk (35%) follow.


Increasing risks for enterprises running on legacy systems report


Why companies hesitate to replace such systems.

The hesitation has one source. The company cannot confirm that new software will adapt to its machines and its operations hardware.

  • Machines expect replies within milliseconds. The company doubts that new software will answer inside that window.

  • Older controllers may use serial connections instead of Ethernet networks. They may also speak older protocols.

  • Cutover, the switch from the old system to the new one. It happens on a live line. A failed switch stops production.

But with legacy system often come some risks

A legacy system is production software that still runs a line but was built for older hardware and older methods. It carries 2 problems.

One, it fails in ways that are slow to notice and costly to fix.

And two, it also leaves the company unsure whether any replacement will work with its machines.

But, with an aging systems like these, risks like the following are commonly found.

  1. Parts run out: Vendors discontinue controllers and spare boards. A failed part then has to come from a resale market or a repair shop.

  2. Support ends: The vendor stops releasing fixes. A station still running it has had no patches since that date.

  3. Knowledge sits with a few people: A very few, or at times one or two engineers know how the system behaves. When they leave, the plant loses that knowledge.

  4. Data stays locked in: Old systems store records in formats that new reporting tools cannot read. Tracking and traceability then depend on manual work.

A digital modernization does not necessarily mean replacing your current systems

A company can add new software around the old system and move one piece at a time. The old system keeps running while each piece is tested.

Strategic workflows like the following make it easy for companies to transform digitally without breaking their ongoing processes.

  1. Observe first

  2. Add a gateway

  3. Run both systems in parallel

The following diagram explains this in further detail


How to transform your systems digitally without breaking the current systems

What does a gateway change?

Let us understand this with an example. A station expects a validation reply within 50ms. The network adds 80ms of delay on one request. We walk through that request.

  1. The station sends a validation request at t = 0ms.

  2. The gateway sits on the machine network. It answers from local rules in 5ms.

  3. The station receives its reply at t = 5ms and continues.

  4. The gateway forwards the data to the new application.

  5. The data arrives 80ms later because of the delay.

  6. The application records it. No machine is waiting.

The input is an 8ms delay. The output is a running line. The slow network no longer matters to the station, because the machine never waits on it.

Bench testing before cutover

A bench test connects the new software to a real controller, away from the line. The test replays recorded traffic. It includes delays of 100ms and dropped messages.

The team confirms that no station would fault. A fault found on the bench costs hours of engineering time. The same fault found on the line stops production.

A phased approach: How the work runs on the shop floor

A phased approach connects new software to a production line one architectural layer at a time. Each layer has its own timing rules, its own protocols and its own owners.

The lower the layer, the stricter the timing. A fault at the lowest layers stops machines, so the phases move from passive observation to live control in 4 steps.

Where Does New Software Sit?

A plant runs on 5 layers. ISA-95 (an international standard for linking plant systems to enterprise systems) numbers them from Level 0 to Level 4.



New software usually arrives at Level 3 or Level 4. The machines sit at Levels 0 and 1. The software has to reach down through Levels 2 and 1 to read data or write results.

The rule that governs every phase is simple. A call from an upper layer must never add a wait to a lower layer.

The Phase 2 and Network Boundaries content now sits inside the 4 phase subsections. The animation, the layer terms and the glance table stay as they are. The text below replaces everything from Phase 1 onward.

Phase 1: Observe and map

In this phase, a network tap (a device that copies network traffic without changing it) or a mirrored switch port records the traffic between PLCs, HMIs and existing systems.


Observing and mapping the current legacy systems


The recording runs long enough to cover every product variant and every shift pattern.

The recording yields 4 outputs:

  1. Protocol inventory: Industrial protocols include Modbus TCP, EtherNet/IP, PROFINET and Siemens S7.

  2. Tag inventory: A tag is a named data point in a PLC, with an address, a data type and a direction.

  3. Handshake map: A handshake is a fixed exchange of signal bits between a PLC and the software above it.

  4. Timing sheet: It lists the response window of each handshake, including the limit set by the watchdog timer.

Old systems also carry behavior that no document records. A fixed wait, a retry rule or a data format quirk can sit in code tuned over many years.

The recording exposes these rules, because the old system still follows them on the wire.

Phase 2: Build the integration layer

In this phase we build the layer between the control level and everything above it. This layer is an edge gateway, an industrial computer placed on the station network. It has 4 jobs:



  1. Protocol translation: It speaks the PLC's native protocol on one side and an IT-friendly protocol on the other, such as OPC UA (a vendor-neutral industrial data standard) or MQTT (a lightweight messaging protocol).

  2. Local state: It holds the rules and reference data needed to answer a PLC without an upstream call.

  3. Store and forward: It queues records on disk while the upstream link is down.

  4. Time handling: It stamps each record with the time of the event, not the time of arrival.

The gateway answers the PLC first and calls the MES afterward. A slow link then delays the record and never the station.

A bench test on a real controller proves this. The team replays recorded traffic, then injects added delay, dropped packets and a full link loss. The watchdog must never trip.

A demilitarized zone (DMZ, a buffer network between plant and corporate networks) keeps corporate systems from connecting to a PLC directly.

Phase 3: Shadow run

A shadow run connects the gateway to the live station in read-only mode. The old system still controls the station.



The gateway receives the same requests and computes its own answers. It writes nothing to the PLC.

A read-only account on the PLC, or a network rule that blocks write commands, enforces this. A software promise to stay read-only is not enough, because a bug can break it.

The team compares the gateway's answer with the old system's answer on every cycle.

Each mismatch traces to 1 of 3 sources: the gateway rules, the reference data or the timing. The team fixes that source and restarts the count of matching cycles from 0.

Phase 4: Staged Cutover

Phase 4 moves control to the gateway, one station at a time. 4 mechanisms keep the move reversible:


Stations cut over one at a time, each reversible to hot standby within one scan cycle.


  1. A control bit in the PLC selects which system answers. An operator sets it from the HMI.

  2. The old system stays in hot standby. It receives the same data and remains ready to answer.

  3. Rollback is the return to the old system. Clearing the control bit takes effect within 1 scan cycle. The team tests this before each cutover.

  4. The next station restarts at Phase 3. Its tags, timing and variants differ from the first station.

What to ask before choosing a team

A partner check is a short list of questions a buyer asks a software team before signing. The reason is the boundary where new application code meets machine controls.

A team that treats a machine like a web service fails at that boundary, and the failure stops a station.

The 5 questions below test for plant-floor knowledge. Each answer can be checked against a document or a demo.

The five questions

A programmable logic controller (PLC) is the industrial computer that runs a station. A protocol is an agreed set of rules for exchanging data.

A watchdog timer is a timer that raises a fault when an expected signal arrives late.


Question

A sound answer contains

A warning sign

Who on the team can read PLC logic?

Named engineers who read ladder logic (the graphical language PLCs use) and tag lists

"The controls team will handle that," with no name

Which industrial protocols has the team shipped?

Named protocols, such as Modbus TCP, EtherNet/IP, PROFINET or OPC UA, with the PLC makes used

A single answer: "REST API integration"

How does the design meet machine timing?

Reply deadlines taken from a recorded timing sheet, and replies sent from local state

Retries and longer timeouts as the plan

Where does the team test?

A bench test on a real controller of the same model, with injected faults, then a read-only run on the live line

Testing only in a simulator or a cloud environment

How does a station return to the old system?

A control bit, a hot standby (the old system receiving the same data) and a rollback tested before each cutover

"Restore from backup"

Software built for your operational reality

Pavesoft Technologies is a manufacturing software partner that builds the software systems that run the shop floor. Our work sits where application code meets machine controls.

What do we bring to the table?

  1. Domain fluency: Our engineers already read PLC logic, protocols and timing sheets. Translation cost is the time and money spent teaching developers how a plant works. We do not bill you for it.

  2. A fluid stack: The machine sets the limits, and our software adapts to them. Delivery runs in 4 to 5 weekly sprints.

  3. Outcome-based delivery: We stay until each phase passes its exit condition.

Our partnership with Paveway Automation

Paveway Automation is a floor automation company that works with automotive manufacturers. It has 15+ years of experience and 300+ projects delivered.



Our formal integration partnership gives each side a clear role. Paveway covers the physical floor. Pavesoft covers the software and data layer.

Together, we deliver 3 things:

  1. A line where hardware and software are planned together.

  2. Software that meets machine timing from the first phase.

  3. Traceability records that start at the station and reach the operations system.

A modernization project earns trust at its first step

In the first step, Pavesoft records your existing traffic and returns a timing sheet for your controls engineer to sign.

Your team then knows how the plant behaves before any code goes live. That record stays with you, whichever team builds next.

Our engineers already speak the language of your floor, and we stay until the results are delivered.

If your plant is weighing a modernization, a Phase 1 assessment is a place to start.

Talk to our experts to scope a realistic path forward.

What does it cost to keep running on a legacy system?

A 2026 survey by TXOne Networks covered about 550 senior decision-makers in 6 European countries. It found 3 things:

  • 60% say legacy Windows makes up at least half of their OT environment.

  • 51% had a security incident involving legacy Windows in the past 12 months. The prior year was 43%.

  • The top barrier to replacement is compatibility with legacy equipment (54%). Cost (38%) and downtime risk (35%) follow.


Increasing risks for enterprises running on legacy systems report


Why companies hesitate to replace such systems.

The hesitation has one source. The company cannot confirm that new software will adapt to its machines and its operations hardware.

  • Machines expect replies within milliseconds. The company doubts that new software will answer inside that window.

  • Older controllers may use serial connections instead of Ethernet networks. They may also speak older protocols.

  • Cutover, the switch from the old system to the new one. It happens on a live line. A failed switch stops production.

But with legacy system often come some risks

A legacy system is production software that still runs a line but was built for older hardware and older methods. It carries 2 problems.

One, it fails in ways that are slow to notice and costly to fix.

And two, it also leaves the company unsure whether any replacement will work with its machines.

But, with an aging systems like these, risks like the following are commonly found.

  1. Parts run out: Vendors discontinue controllers and spare boards. A failed part then has to come from a resale market or a repair shop.

  2. Support ends: The vendor stops releasing fixes. A station still running it has had no patches since that date.

  3. Knowledge sits with a few people: A very few, or at times one or two engineers know how the system behaves. When they leave, the plant loses that knowledge.

  4. Data stays locked in: Old systems store records in formats that new reporting tools cannot read. Tracking and traceability then depend on manual work.

A digital modernization does not necessarily mean replacing your current systems

A company can add new software around the old system and move one piece at a time. The old system keeps running while each piece is tested.

Strategic workflows like the following make it easy for companies to transform digitally without breaking their ongoing processes.

  1. Observe first

  2. Add a gateway

  3. Run both systems in parallel

The following diagram explains this in further detail


How to transform your systems digitally without breaking the current systems

What does a gateway change?

Let us understand this with an example. A station expects a validation reply within 50ms. The network adds 80ms of delay on one request. We walk through that request.

  1. The station sends a validation request at t = 0ms.

  2. The gateway sits on the machine network. It answers from local rules in 5ms.

  3. The station receives its reply at t = 5ms and continues.

  4. The gateway forwards the data to the new application.

  5. The data arrives 80ms later because of the delay.

  6. The application records it. No machine is waiting.

The input is an 8ms delay. The output is a running line. The slow network no longer matters to the station, because the machine never waits on it.

Bench testing before cutover

A bench test connects the new software to a real controller, away from the line. The test replays recorded traffic. It includes delays of 100ms and dropped messages.

The team confirms that no station would fault. A fault found on the bench costs hours of engineering time. The same fault found on the line stops production.

A phased approach: How the work runs on the shop floor

A phased approach connects new software to a production line one architectural layer at a time. Each layer has its own timing rules, its own protocols and its own owners.

The lower the layer, the stricter the timing. A fault at the lowest layers stops machines, so the phases move from passive observation to live control in 4 steps.

Where Does New Software Sit?

A plant runs on 5 layers. ISA-95 (an international standard for linking plant systems to enterprise systems) numbers them from Level 0 to Level 4.



New software usually arrives at Level 3 or Level 4. The machines sit at Levels 0 and 1. The software has to reach down through Levels 2 and 1 to read data or write results.

The rule that governs every phase is simple. A call from an upper layer must never add a wait to a lower layer.

The Phase 2 and Network Boundaries content now sits inside the 4 phase subsections. The animation, the layer terms and the glance table stay as they are. The text below replaces everything from Phase 1 onward.

Phase 1: Observe and map

In this phase, a network tap (a device that copies network traffic without changing it) or a mirrored switch port records the traffic between PLCs, HMIs and existing systems.


Observing and mapping the current legacy systems


The recording runs long enough to cover every product variant and every shift pattern.

The recording yields 4 outputs:

  1. Protocol inventory: Industrial protocols include Modbus TCP, EtherNet/IP, PROFINET and Siemens S7.

  2. Tag inventory: A tag is a named data point in a PLC, with an address, a data type and a direction.

  3. Handshake map: A handshake is a fixed exchange of signal bits between a PLC and the software above it.

  4. Timing sheet: It lists the response window of each handshake, including the limit set by the watchdog timer.

Old systems also carry behavior that no document records. A fixed wait, a retry rule or a data format quirk can sit in code tuned over many years.

The recording exposes these rules, because the old system still follows them on the wire.

Phase 2: Build the integration layer

In this phase we build the layer between the control level and everything above it. This layer is an edge gateway, an industrial computer placed on the station network. It has 4 jobs:



  1. Protocol translation: It speaks the PLC's native protocol on one side and an IT-friendly protocol on the other, such as OPC UA (a vendor-neutral industrial data standard) or MQTT (a lightweight messaging protocol).

  2. Local state: It holds the rules and reference data needed to answer a PLC without an upstream call.

  3. Store and forward: It queues records on disk while the upstream link is down.

  4. Time handling: It stamps each record with the time of the event, not the time of arrival.

The gateway answers the PLC first and calls the MES afterward. A slow link then delays the record and never the station.

A bench test on a real controller proves this. The team replays recorded traffic, then injects added delay, dropped packets and a full link loss. The watchdog must never trip.

A demilitarized zone (DMZ, a buffer network between plant and corporate networks) keeps corporate systems from connecting to a PLC directly.

Phase 3: Shadow run

A shadow run connects the gateway to the live station in read-only mode. The old system still controls the station.



The gateway receives the same requests and computes its own answers. It writes nothing to the PLC.

A read-only account on the PLC, or a network rule that blocks write commands, enforces this. A software promise to stay read-only is not enough, because a bug can break it.

The team compares the gateway's answer with the old system's answer on every cycle.

Each mismatch traces to 1 of 3 sources: the gateway rules, the reference data or the timing. The team fixes that source and restarts the count of matching cycles from 0.

Phase 4: Staged Cutover

Phase 4 moves control to the gateway, one station at a time. 4 mechanisms keep the move reversible:


Stations cut over one at a time, each reversible to hot standby within one scan cycle.


  1. A control bit in the PLC selects which system answers. An operator sets it from the HMI.

  2. The old system stays in hot standby. It receives the same data and remains ready to answer.

  3. Rollback is the return to the old system. Clearing the control bit takes effect within 1 scan cycle. The team tests this before each cutover.

  4. The next station restarts at Phase 3. Its tags, timing and variants differ from the first station.

What to ask before choosing a team

A partner check is a short list of questions a buyer asks a software team before signing. The reason is the boundary where new application code meets machine controls.

A team that treats a machine like a web service fails at that boundary, and the failure stops a station.

The 5 questions below test for plant-floor knowledge. Each answer can be checked against a document or a demo.

The five questions

A programmable logic controller (PLC) is the industrial computer that runs a station. A protocol is an agreed set of rules for exchanging data.

A watchdog timer is a timer that raises a fault when an expected signal arrives late.


Question

A sound answer contains

A warning sign

Who on the team can read PLC logic?

Named engineers who read ladder logic (the graphical language PLCs use) and tag lists

"The controls team will handle that," with no name

Which industrial protocols has the team shipped?

Named protocols, such as Modbus TCP, EtherNet/IP, PROFINET or OPC UA, with the PLC makes used

A single answer: "REST API integration"

How does the design meet machine timing?

Reply deadlines taken from a recorded timing sheet, and replies sent from local state

Retries and longer timeouts as the plan

Where does the team test?

A bench test on a real controller of the same model, with injected faults, then a read-only run on the live line

Testing only in a simulator or a cloud environment

How does a station return to the old system?

A control bit, a hot standby (the old system receiving the same data) and a rollback tested before each cutover

"Restore from backup"

Software built for your operational reality

Pavesoft Technologies is a manufacturing software partner that builds the software systems that run the shop floor. Our work sits where application code meets machine controls.

What do we bring to the table?

  1. Domain fluency: Our engineers already read PLC logic, protocols and timing sheets. Translation cost is the time and money spent teaching developers how a plant works. We do not bill you for it.

  2. A fluid stack: The machine sets the limits, and our software adapts to them. Delivery runs in 4 to 5 weekly sprints.

  3. Outcome-based delivery: We stay until each phase passes its exit condition.

Our partnership with Paveway Automation

Paveway Automation is a floor automation company that works with automotive manufacturers. It has 15+ years of experience and 300+ projects delivered.



Our formal integration partnership gives each side a clear role. Paveway covers the physical floor. Pavesoft covers the software and data layer.

Together, we deliver 3 things:

  1. A line where hardware and software are planned together.

  2. Software that meets machine timing from the first phase.

  3. Traceability records that start at the station and reach the operations system.

A modernization project earns trust at its first step

In the first step, Pavesoft records your existing traffic and returns a timing sheet for your controls engineer to sign.

Your team then knows how the plant behaves before any code goes live. That record stays with you, whichever team builds next.

Our engineers already speak the language of your floor, and we stay until the results are delivered.

If your plant is weighing a modernization, a Phase 1 assessment is a place to start.

Talk to our experts to scope a realistic path forward.

“Before automation, scaling our operations felt chaotic. Now everything runs in the background — and we focus on growth instead of maintenance”

FAQ's

What is translation cost?

What is translation cost?

Why can’t template software fit a manufacturing process?

Why can’t template software fit a manufacturing process?

How long does it take to reach go-live?

How long does it take to reach go-live?

How do Pavesoft and Paveway Automation work together?

How do Pavesoft and Paveway Automation work together?

Know exactly what you're building before you commit to it.

One discovery call. A clear scope, a realistic timeline, and no obligation.

Know exactly what you're building before you commit to it.

One discovery call. A clear scope, a realistic timeline, and no obligation.

Know exactly what you're building before you commit to it.

One discovery call. A clear scope, a realistic timeline, and no obligation.

© 2026 Pavesoft Technologies. All rights reserved.

© 2026 Pavesoft Technologies. All rights reserved.