Connected Machine Architecture for Legacy and New Equipment

The reshoring boom is transforming American manufacturing. Record commitments totaling 1.9 million announced jobs since 2010 are driving unprecedented greenfield construction — from semiconductor fabs funded by the CHIPS Act to EV battery plants, aerospace expansion, and defense manufacturing buildouts. But here's the reality that every new US factory faces: greenfield doesn't mean all-new equipment. Companies transfer proven legacy machines from existing facilities. They acquire used CNC centers, presses, and grinders because lead times for new equipment stretch 12-18 months. They inherit equipment from acquisitions. The result is a factory floor where a brand-new Siemens S7-1500 with native OPC-UA sits next to a 1990s Fanuc 16i-MB with serial RS-232 and a 2005 Allen-Bradley SLC 500 running Modbus. In twenty years of designing connected factories across the US, I've never seen a greenfield facility that was 100% new equipment — the typical mix is 50-70% new machines and 30-50% transferred or acquired legacy assets. The challenge is making them all look the same to your analytics platform. A spindle temperature reading from a 2024 DMG MORI with OPC-UA should be indistinguishable from a spindle temperature reading from a 1995 Haas VF-2 with a retrofit thermocouple — same data format, same timestamp precision, same quality tags, same dashboard. We design connected machine architectures that bridge every generation of equipment into a unified data model: native OPC-UA extraction from modern machines, retrofit sensors and protocol gateways for legacy equipment, and edge compute that makes all machines equal citizens in your AI analytics platform. Schedule a Demo

Two Paths, One Data Model — Every Machine Connected

Modern Machines (50-70%)

OPC-UA server built into PLC MTConnect native support FANUC FOCAS / Mazak SmoothConnect Servo data, cycle times, fault codes

Direct Protocol Extraction

Edge Gateway

Unified Data Model

Legacy Equipment (30-50%)

No network connection Modbus RTU / serial RS-232 Relay logic / hard-wired I/O No digital diagnostics

Retrofit Sensors + Protocol Gateway

The US Manufacturing Reality

$238B

Construction Spending Record

US manufacturing construction spending reached a record $238 billion in 2024 — driven by CHIPS Act semiconductor investments, EV battery gigafactories, and defense manufacturing expansion. Every new facility being built needs machine connectivity architecture designed during construction, not after equipment arrives.

1.9M

Reshoring Jobs Announced

Record reshoring commitments from 2010-2023 are creating new US factories that combine brand-new automation with transferred equipment from closed overseas facilities or consolidated domestic plants. These mixed-generation shops need unified connectivity from day one.

85%

Machines Not Connected

Globally, 85% of factory machines and inventory are not yet connected to the internet. In the US, even new facilities struggle with legacy equipment connectivity because the OEMs who built the machines 15-25 years ago no longer support them — and their protocols died with them.

1.9M

Workforce Gap by 2033

The US manufacturing talent gap may leave 1.9 million positions unfilled by 2033. Connected machine architectures with automated data collection replace manual operator logging — critical when you can't hire enough skilled workers to monitor equipment the old-fashioned way.

Building a new US manufacturing facility with mixed equipment generations? Schedule a demo to see how we design connectivity for both brand-new CNC machines and transferred legacy equipment — unified analytics from commissioning day.

Machine Connectivity Tiers

Tier 5

Fully Connected (2018+)

Native OPC-UA server in PLC. MTConnect agent available. Manufacturer API (FANUC FOCAS, Mazak SmoothConnect, Haas MNET). Full servo data, program status, tool data, cycle times, fault codes, and diagnostics accessible over Ethernet. Connectivity cost: $500-$2,000 (OPC-UA license if not included). Examples: Siemens S7-1500, Rockwell ControlLogix, Beckhoff TwinCAT 3, DMG MORI CELOS, new Mazak/Okuma/Makino.

Tier 4

Network-Ready (2008-2018)

Ethernet port available but no OPC-UA. Proprietary protocols (FANUC FOCAS2, older Siemens S7-300/400 via PUT/GET, Allen-Bradley via EtherNet/IP). Data extractable via protocol-specific gateway (Kepware, Ignition). Connectivity cost: $2,000-$5,000 (gateway + license). Examples: FANUC 30i/31i/32i, Siemens 840D sl, older Rockwell CompactLogix.

Tier 3

Serial / Fieldbus (1995-2008)

RS-232/RS-485 serial port or PROFIBUS/DeviceNet fieldbus. Limited data available: run/stop status, alarm codes, basic counters. Protocol gateway required for Modbus RTU to Ethernet translation. May require PLC program modification to expose additional data points. Connectivity cost: $3,000-$8,000. Examples: FANUC 16i/18i/21i, Allen-Bradley SLC 500, Siemens S7-300 (PROFIBUS only).

Tier 2

Closed Controller (1985-1995)

Proprietary controller with no standard communication interface. May have discrete I/O available (24V signals for run, fault, cycle complete). Data extraction: clamp-on current sensors on motor cables, external relay to capture machine state from indicator lights. Connectivity cost: $5,000-$15,000 (external sensors + edge I/O + gateway). Examples: Older Mazak T-Plus, Cincinnati Milacron Acramatic, legacy injection molding machines.

Tier 1

No Controller / Relay Logic (Pre-1985)

No digital controller — hard-wired relay logic, manual controls, or simple PLC with no communication. All data must come from external sensors: current transformers on motor leads (run/idle/load), vibration accelerometers on bearings, temperature RTDs on critical points, proximity sensors for cycle counting. Connectivity cost: $8,000-$25,000 (full sensor suite + edge I/O + gateway). Examples: Manual lathes with DRO only, older hydraulic presses, legacy special-purpose machines.

Modern Machine: Native Data Extraction

Machine Brand	Protocol	Available Data	Integration Tool	Greenfield Spec
Siemens SINUMERIK 840D/ONE	OPC-UA native	Servo positions, spindle load, tool data, program, alarms, energy	Direct OPC-UA; SINUMERIK Edge	Specify OPC-UA license + Edge device in PO
FANUC 30i/31i/0i-TF+	FOCAS2 API; MT-LINKi	Servo, spindle, PMC, macro, alarm, program, tool management	Kepware FANUC driver; MT-LINKi	Specify FOCAS Ethernet option in PO
Mazak SmoothAi / SmoothX	MTConnect; SmoothConnect	Spindle, axis, tool, program, thermal, energy, maintenance	Built-in MTConnect agent	Verify SmoothConnect enabled at delivery
Haas NGC	MNET (Ethernet); MDC	Machine state, spindle, axis, macro variables, alarms	HaasConnect; Kepware Haas driver	Specify Ethernet + MDC option in PO
Okuma OSP-P300/P500	MTConnect; OPC-UA	Spindle, axis, thermal, program, maintenance, energy	Okuma Connect Plan	Specify Connect Plan license in PO
DMG MORI CELOS	OPC-UA; CELOS Connect	Full machine data, tool management, condition monitoring	CELOS Connectivity; direct OPC-UA	CELOS Connect included — verify activation
Makino ProNet	MTConnect; OPC-UA	Spindle, axis, thermal, program, maintenance	ProNet Connect	Specify ProNet license in PO
KUKA / FANUC / ABB Robots	OPC-UA; proprietary API	Joint torque, temperature, cycle, path accuracy, I/O	KUKA.Connect; FANUC iRVision; ABB OmniCore	Specify OPC-UA server license in PO

Legacy Machine: Retrofit Sensor Strategy

Power & Current Monitoring

Non-invasive split-core current transformers (CTs) clamped around motor power cables — no electrical modification required, no warranty impact. Measures: run/idle/off state, load percentage, power consumption, and motor health indicators. A $50 CT on each major motor (spindle, axis drives, hydraulic pump) provides 80% of the machine state information that a modern PLC exposes natively. This is the single highest-ROI retrofit for legacy machines. Install time: 30 minutes per motor. Zero downtime required.

Vibration Monitoring

External accelerometers mounted on spindle bearing housings, gearbox casings, and motor frames. Stud-mounted (M5/M8 threaded pad) for permanent installation or magnetic mount for assessment. Detects bearing wear, gear mesh issues, imbalance, and misalignment — the same failure modes that modern machine diagnostics report internally. For legacy machines with no built-in diagnostics, external vibration monitoring is the only way to predict mechanical failures. Cost: $150-$500 per sensor point.

Temperature Monitoring

Surface-mount RTDs (PT100) on spindle housings, bearing caps, hydraulic reservoirs, and coolant systems. Detects thermal trends that indicate lubrication degradation, bearing wear, or cooling system problems. For CNC machines with no thermal compensation: correlating ambient and spindle temperature with dimensional accuracy reveals environmental root causes of quality drift. Cost: $30-$100 per point. Wireless (LoRaWAN) temperature sensors eliminate cabling entirely at $50-$150 per point.

Discrete I/O Capture

Many legacy machines have indicator lights, relays, or 24V signals for cycle complete, fault, door open/closed, and spindle running. Tapping these signals with an edge I/O module (Opto 22 groovRIO, Beckhoff EK9160, or similar) captures machine state without modifying the machine's control system. Cycle complete signal → cycle count and cycle time calculation. Fault relay → downtime tracking. Spindle run signal → utilization percentage. Cost: $500-$2,000 for edge I/O module with 8-16 digital inputs.

Acoustic / Ultrasonic

External microphones for tool wear detection (grinding, machining), air leak detection (pneumatic systems), and bearing early-warning (ultrasonic emissions before vibration signature appears). Particularly valuable for legacy machines with no tool wear monitoring — acoustic emission correlates with cutting force and tool condition. Non-contact, non-invasive, and zero modification to the machine. Cost: $200-$1,000 per sensor.

Vision-Based State Detection

AI camera monitoring machine indicator panel (stack lights, HMI screens, analog gauges) to extract state information from machines that have no digital interface whatsoever. A $200 camera pointed at a stack light provides run/idle/fault detection without touching the machine's electrical system. For analog gauges: computer vision reads gauge position and digitizes the reading. This approach sounds unconventional but is increasingly common for machines where any electrical modification is prohibited.

Protocol Translation Architecture

Source

Every Protocol on Your Floor

Modbus RTU (serial RS-485), Modbus TCP, EtherNet/IP (CIP), PROFINET, PROFIBUS, DeviceNet, CC-Link, FANUC FOCAS, Mazak Mazatrol, Haas MNET, Siemens S7, BACnet, MTConnect, analog 4-20mA/0-10V, discrete 24V I/O. A typical US factory with mixed equipment has 5-8 different protocols running simultaneously. Each requires its own driver in the gateway software.

Gateway

Software Gateway: Protocol → OPC-UA/MQTT

Kepware (PTC ThingWorx), Ignition (Inductive Automation), HighByte Intelligence Hub, Softing, or Matrikon OPC — these software gateways support 100+ industrial protocol drivers and expose all connected devices as a unified OPC-UA address space or MQTT publisher. One gateway instance can simultaneously connect to Siemens, Rockwell, FANUC, Modbus, and BACnet devices. In greenfield: gateway software specified during automation design, pre-loaded and configured during commissioning. Gateway hardware: industrial PC or VM on edge server.

Output

Unified OPC-UA / MQTT + Sparkplug B

All machine data — regardless of source protocol — published to the Unified Namespace as standardized OPC-UA tags or Sparkplug B MQTT payloads. A temperature reading from a 2024 Siemens S7-1500 via native OPC-UA and a temperature reading from a 1995 Haas VF-2 via retrofit RTD + Modbus gateway appear identical in the analytics platform: same topic structure, same data type, same engineering units, same timestamp precision, same quality code. The consuming application cannot tell — and does not need to know — which machine is new and which is legacy.

Unified Data Model Design

Machine State (Every Machine)

Standardized machine state enumeration: Running, Idle, Fault, Setup/Changeover, Planned Downtime, Unplanned Downtime. For modern machines: state derived from PLC program status and servo enable signals via OPC-UA. For legacy machines: state derived from motor current (running vs idle), fault relay (fault), and operator input on HMI touchscreen (setup/changeover). Both paths produce identical state transitions in the analytics platform — enabling OEE calculation across all machines regardless of generation.

Cycle Data (Every Machine)

Cycle start timestamp, cycle end timestamp, cycle time (seconds), parts count increment, program/recipe identifier. Modern machines: extracted from PLC cycle complete signal and program counter via OPC-UA. Legacy machines: cycle complete from discrete I/O (proximity sensor on part ejector, relay on press bottom-dead-center, or current signature cycle detection). Both paths feed the same cycle time analytics, production rate tracking, and efficiency calculations.

Health Data (Tier-Dependent)

Vibration RMS, bearing temperature, spindle load, coolant flow, hydraulic pressure. Modern machines: full diagnostic dataset from PLC including servo following error, drive temperature, and internal fault codes. Legacy machines: subset from retrofit sensors — vibration, temperature, and current provide the core predictive maintenance signals. The data model accommodates both complete and partial health datasets — analytics models trained per machine tier, not per machine brand.

Quality Integration

Inspection results linked to machine, cycle, and process parameters for root cause analysis. Modern machines: process parameters (spindle speed, feed rate, tool number) automatically tagged to each part. Legacy machines: cycle timestamp correlated with inspection data by time alignment — parts produced between timestamp A and B assigned the quality results from that batch. Both approaches enable the same correlation analysis: "which machine conditions produce out-of-spec parts?"

Need a unified data model that treats legacy and modern machines identically? Schedule a demo to see how our connected machine architecture delivers one dashboard, one analytics platform, and one AI engine across every generation of equipment on your floor.

Key Benefits & ROI

100% Connectivity — every machine, every generation, every brand connected

1 Platform Single analytics — legacy and modern machines on one dashboard

$0 Machine modification — retrofit sensors without touching controls

Standard Data model — OPC-UA/MQTT for all, regardless of source protocol

Upgrade Path — legacy machines graduate to higher tiers as controls are replaced

Your 1995 Haas Should Look Identical to Your 2024 DMG MORI

iFactory designs connected machine architectures for US greenfield factories — native OPC-UA from modern equipment, retrofit sensors for legacy machines, protocol translation for everything in between — so your AI analytics platform sees one unified data model from commissioning day.

Schedule a Demo Talk to Support

Frequently Asked Questions

How do you connect machines without OPC-UA?

Three approaches depending on what the machine has. (1) If the machine has any Ethernet port and a proprietary protocol (FANUC FOCAS, Haas MNET, older Siemens S7): use a software gateway like Kepware or Ignition with the vendor-specific driver. These gateways have drivers for 100+ protocols and expose data as OPC-UA or MQTT. (2) If the machine has serial communication (RS-232/RS-485 Modbus RTU, older CNC DNC ports): use a serial-to-Ethernet converter paired with a Modbus TCP gateway. Limited data may be available — typically run/stop status, alarm codes, and basic counters. (3) If the machine has no digital communication at all: use external sensors. Non-invasive current transformers on motor cables detect run/idle/load state. Discrete I/O tapping captures cycle complete and fault signals. Retrofit vibration and temperature sensors provide health data. The edge I/O module (Opto 22 groovRIO, Beckhoff EK9160) digitizes all signals and publishes to MQTT. In all three cases, the data enters the same Unified Namespace — consuming applications can't distinguish the source.

What retrofit sensors work on old CNC machines?

The highest-ROI retrofit package for a legacy CNC is: (1) Split-core current transformers on spindle motor and axis drive cables — $50-$200 per CT, zero downtime installation, provides run/idle/fault state, load percentage, and power consumption. (2) Accelerometer on spindle bearing housing — $150-$500, stud-mounted for permanent monitoring, detects bearing wear 4-8 weeks before failure. (3) Surface RTD on spindle housing — $30-$100, monitors thermal behavior for dimensional accuracy correlation. (4) Proximity sensor or signal tap on cycle complete relay — $50-$200, provides cycle count and cycle time for OEE calculation. Total cost: $500-$2,000 per machine. Installation: 2-4 hours during a planned maintenance window. No modification to the machine's electrical system — all sensors are external, non-invasive. This package delivers 80% of the data value that a fully connected modern machine provides, at 10% of the cost of a machine replacement.

How do you extract data without voiding the warranty?

Two principles: read-only access and external-only sensors. For modern machines with OPC-UA or manufacturer APIs: configure the gateway for read-only access. OPC-UA clients can subscribe to data without writing back to the PLC. FANUC FOCAS2 can be configured in monitor-only mode. Read-only data extraction does not modify the machine's control program or parameters — the warranty remains intact. For legacy machines: all retrofit sensors are external to the machine's electrical system. Current transformers clamp around cables without cutting or modifying them. Vibration accelerometers mount on the outside of housings. Temperature RTDs attach to surfaces with thermal paste or clamps. No wires are connected to the machine's control cabinet, no PLC programs are modified, no electrical connections are altered. CNC OEMs including Mazak, Okuma, and DMG MORI explicitly support third-party monitoring via their APIs and have published integration guides — using these documented interfaces does not void warranty.

How do you convert Modbus to MQTT?

A software gateway running on an edge device (industrial PC, Raspberry Pi for testing, or VM) connects to Modbus RTU devices via serial port or Modbus TCP devices via Ethernet. The gateway reads specified Modbus registers on a polling schedule (typically 1-10 second intervals), maps each register to a named tag with engineering units, and publishes changes to an MQTT broker as Sparkplug B payloads. Practical example: Modbus register 40001 (holding register 1) contains spindle speed in RPM. The gateway reads this register every second, converts the raw integer to RPM, and publishes to topic spBv1.0/plant/DDATA/area/machine_01 with metric name "spindle_rpm", data type Float, and timestamp. Tools: Kepware Modbus driver, Ignition Modbus module, Node-RED with node-red-contrib-modbus, or Python with pymodbus + paho-mqtt. In greenfield, this translation is configured during commissioning and runs autonomously — no ongoing engineering required.

How do you standardize data from different machine brands?

Through the unified data model layer. Regardless of source protocol, every machine's data is mapped to a standard schema: machine state (enum: Running/Idle/Fault/Setup), cycle data (start time, end time, duration, parts count), health data (vibration, temperature, current, pressure with engineering units), and identity (machine ID, location, type, manufacturer). This mapping happens in the edge gateway — Kepware, Ignition, or HighByte map vendor-specific tag names to standard names. A Siemens PLC might call spindle speed "DB100.DBW0"; a FANUC calls it via FOCAS "Spindle Speed Actual"; a Modbus register is just "40001." All three are mapped to the standard tag "spindle_speed_rpm" in the unified namespace. The analytics platform subscribes to "spindle_speed_rpm" without knowing or caring which brand produced the data. OPC-UA companion specifications (PackML, umati for machine tools) provide standardized information models that eliminate custom mapping for supported machine types. Schedule a demo to see the unified data model across mixed equipment fleets.

Greenfield + Legacy = Connected. Not Compromised.

50-70% new machines with OPC-UA. 30-50% transferred legacy equipment with retrofit sensors. 100% connected to one analytics platform. That's the reality of American manufacturing in 2026 — and we design for it.

Schedule a Demo Talk to Support