top of page
Smartic Corporation
Search

Industrial Retrofit: How to Connect 20-Year-Old Machinery Without Replacing It

Nov 4
7 min read

The Objection That Halts Half of All Projects


"Our machines are old; there's no way to connect them."


This is the single phrase that kills digitalization projects before they even start, resting on a flawed premise: that Industry 4.0 begins with buying brand-new machinery.


Schneider Electric transformed its Lexington, Kentucky plant—a facility with over 60 years of operation—into its first smart factory in the United States. They didn't replace it; they instrumented it. Published results show 3.5% annual energy savings, a 20% reduction in MTTR for critical equipment, a 5% reduction in unplanned downtime in critical processes, and $500,000 in work-in-process inventory eliminated through end-to-end traceability.


A six-decade-old plant can generate world-class data. What's required isn't new steel: it's a layer of sensing, data acquisition, and communication mounted on the steel you already own.


Retrofit and Replacement Are Not Competitors: They Solve Different Problems


Machine Replacement

Retrofit

What Changes

The entire physical asset

Instrumentation and connectivity are added

Investment

High CAPEX

Marginal investment relative to asset value

Downtime

Long: dismantling, assembly, commissioning

Minimal to zero: mostly installed while operating or during scheduled stops

Operational Risk

Process re-qualification, operator learning curve

Minimal if external sensing is used

When It Makes Sense

Mechanical structure reached end-of-life or process changed

Mechanics remain capable, but the machine is "mute"


The key question to structure the decision is simple: Is this machine's problem mechanical or informational? If it can no longer hold tolerances, no sensor will fix it. If the problem is that nobody knows how often it stops, why it stops, or how much power it consumes, it's an information problem—and that is infinitely cheaper to solve.


The Cost of Inaction


The True Cost of Downtime study by Siemens/Senseye quantifies the loss: the world's largest industrial companies lose approximately $1.4 trillion annually due to unplanned downtime, equivalent to 11% of their total revenue. An average plant suffers roughly 326 hours of unplanned downtime per year.


However, the most revealing insight is this: between the two most recent studies, downtime incidents dropped from 42 to 25 per month, while recovery time per incident jumped from 49 to 81 minutes. In short: failures occur less frequently, but they are harder to diagnose. And diagnosing without data is simply guessing faster, not guessing better.


Deloitte reinforces this from a capacity standpoint: a deficient maintenance strategy reduces plant production capacity by 5% to 20%. That is capacity you have already purchased, already financed, and are currently wasting.


How many hours did your primary line stop last month, and why? If the answer is in a supervisor's memory rather than a system, that is precisely the problem.

We can conduct an assessment during a 2-hour site visit. [Request Site Visit]


How to Instrument a Machine Without a PLC


The secret lies in non-invasive sensing: instrumentation that does not touch control circuits, modify existing code, or require opening high-voltage cabinets.


  1. Accelerometers (Vibration): Magnetic or stud-mounted onto bearing housings. They detect unbalance, misalignment, looseness, bearing degradation, and cavitation. The evaluation framework is ISO 20816-3:2022 for machines over 15 kW between 120 and 30,000 r/min. It represents the best signal-to-cost ratio sensor for critical rotating assets: motors, pumps, fans, compressors, and gearboxes.

  2. Clamp-on Current Transformers (Non-Invasive CTs): Clamped around phase conductors without cutting wires or de-energizing equipment. They provide RMS current and, with known voltage, estimated power. This is the most underrated sensor in retrofit projects: an electrical signature reveals full operational state without PLC data—clearly distinguishing powered-off, starting, idling, loaded, and jammed states. It is the most cost-effective path to calculate availability and OEE on completely mute machines.

  3. Temperature Sensors: Surface PT100s, thermocouples, or non-contact infrared sensors for rotating or inaccessible points. They monitor bearings, motor casings, oil levels, glands, ovens, and chambers.

  4. Pulse Counters and Discrete Sensors: Inductive, photoelectric, or reed sensors mounted on elements moving once per cycle (cam, cylinder, ejector, line exit) turn mechanical motion into digital counts. One pulse = one unit produced. This enables the plant's most actionable metric: kWh per unit produced. Combined with a CT, any 1990s machine yields specific energy efficiency.

  5. Computer Vision: A camera aimed at the stack light (Andon tower) reads machine status without electrical contact. Ideal when OEMs prohibit control panel access or when equipment lacks digital outputs. Also useful for unit counting and inline inspection.


Illustrating the power of external instrumentation, ABB reports that its wireless retrofit sensor for low-voltage motors—applicable to any brand and installed "in minutes without additional wiring"—reduces downtime by up to 70%, extends motor lifespan by up to 30%, and lowers energy consumption by up to 10%.


The Architecture: From Sensor to Decision


Data does not travel on its own. Below is the operational chain and underlying protocols:


  • Modbus RTU: Serial over RS-485, master/slave. The lowest common denominator in legacy instrumentation: nearly every VFD, power meter, or temperature controller from the last twenty years speaks it.

  • Modbus TCP: The same data model over Ethernet, port 502.

  • PROFINET: Deterministic industrial Ethernet, relevant when legacy equipment includes modern PLCs.

  • OPC UA: The interoperability layer. Its real value lies in information modeling: instead of "Register 40001", data is structured as Line3.Extruder.Motor1.Temperature. Released in 2008, it runs on PCs, cloud servers, PLCs, and ARM microcontrollers with native security—X.509 certificates, message signing, encryption, and auditing.

  • MQTT: Lightweight publish/subscribe protocol over TCP, engineered for intermittent, low-bandwidth links. An OASIS standard since March 2019 (v5.0), it transports plant data to cloud platforms without opening inbound ports on OT networks.

  • Edge Gateway: The translator and boundary device. It reads fieldbus sensors, normalizes data, executes local logic, buffers data during network outages (store-and-forward), and publishes upstream. It establishes physical and logical segmentation between OT and IT networks.


Need a reference architecture diagram for a brownfield plant?

We can send it to you in PDF format, complete with a recommended sensor checklist by asset type. [Request Diagram]


The Four Real Risks (and How to Neutralize Them)


  1. Voiding OEM Warranties

    Intervening in control cabinets, power wiring, or firmware can void manufacturer warranties. Mitigation: Use strictly external sensing—clamp-on CTs, magnetic accelerometers, non-contact infrared, and vision systems. Require zero modifications to control panels or PLC code, and document approval with the OEM prior to installation.


  2. Exposing OT Networks

    Connecting a machine isolated for twenty years introduces cyber vulnerabilities. The governing standards are NIST SP 800-82 Rev. 3 (Guide to Operational Technology Security, Sep 2023) and the ISA/IEC 62443 series—recognized by ISA as the consensus cybersecurity standard for automation and control systems. Plant managers should reference ANSI/ISA-62443-2-1 for security program requirements; integrators apply 62443-2-4; product vendors follow 62443-4-1 and 4-2. Core principle: Gateways publish outbound only; zero inbound ports are opened into OT networks.


  3. Electrical Noise

    VFDs, soft starters, and welding equipment induce electrical noise onto analog signals and RS-485 buses. Mitigation: Use shielded twisted-pair cabling grounded at a single end, maintain physical separation from power trays, ensure proper bus termination/biasing, utilize galvanic isolation on gateway inputs, and prefer 4–20 mA current loops over voltage signals. (This is the #1 cause of projects that work in trials but fail in production).


  4. Data Without Context

    A sensor unlinked to baseline metrics or production events produces noise, not actionable intelligence. Before installing, define the precise business question each sensor must answer.


The Seven Steps of a Retrofit Project


  1. Asset Inventory & Criticality Mapping: Identify bottleneck machinery and calculate hourly downtime costs. Translate Siemens' 11% downtime revenue impact into your plant's balance sheet.

  2. Define Questions Before Sensors: Are you measuring Availability, Specific Consumption, or Quality? Goal definitions dictate sensor selection. Installing sensors without clear objectives creates dashboards nobody uses.

  3. Single-Machine Bottleneck Pilot: Deploy non-invasive sensing, an edge gateway, and a dashboard on one critical asset within weeks, not months.

  4. Validation Against Operator Reality: Reconcile digital data with shift logs before making management decisions. Discrepancies reveal sensor calibration needs—or inaccuracies in manual shift reporting.

  5. Communication & Security Architecture Definition: Establish OT/IT segmentation, enforce outbound-only gateway rules, and apply NIST 800-82r3 and IEC 62443 standards.

  6. Multi-Line Scaling with Common Data Models: Leverage OPC UA information models across assets to standardize integration rather than re-engineering per machine.

  7. Integration with Maintenance & Energy Systems: Automate work order generation and feed energy management KPIs. Data that doesn't drive action is mere decoration.


An Honest Note on Costs


No independent, reputable source publishes universal cost ranges for industrial retrofit projects: neither the DOE, IEA, nor equipment manufacturers release standardized pricing. Any arbitrary figures found online represent, at best, single-case anecdotal data.


However, validated return metrics compiled by the U.S. Department of Energy's Industrial Assessment Centers (and systematized by ACEEE) demonstrate that implemented efficiency measures achieve payback in under one year on average, generating approximately $44,000 in annual savings per facility.


When evaluating vendors, request the pilot cost for one machine alongside the marginal cost for the tenth machine. If the tenth machine isn't substantially cheaper, you are being sold ten bespoke projects rather than a scalable platform.


Conclusion


Industry 4.0 does not start in a new machinery catalog. It starts with the questions your plant cannot answer today due to a lack of data.


A mechanically sound 20-year-old machine is a prime retrofit candidate: the mechanical asset is fully amortized, operators understand it, and processes are qualified. It simply lacks a voice. Giving it a voice costs a fraction of total replacement—and when replacement eventually becomes necessary, you will possess exact operational data to justify your next acquisition.


Have a critical machine operating in the dark? Let's run a pilot on that asset.

Schedule a 30-minute session to define project scope and requirements. [Schedule Session]

Referencias y fuentes

Article prepared for Smartic — Industry 4.0, Industrial IoT, and Predictive Maintenance. Want to see Smartic in action at your plant? Schedule a 30-minute demo.

 
 
 

Comments


© 2022 by Smartic Corporation

bottom of page