What Is a Machine Monitoring System? How to Start Collecting Machine Data

Many plants know their machines "stop a lot", but can't say for how long, why, or which machine is the real bottleneck, because the data comes from paper logs filled in at the end of the shift. A machine monitoring system captures this information directly from the equipment. This article explains how these systems work, what data is worth collecting, whether older machines can be connected, and how to start without wasting money.

What is a machine monitoring system, and how is it different from manual logs?

A machine monitoring system connects to your machines and records their operating state automatically: running, waiting for parts, stopped, or in alarm. The data is sent to a server or central database and presented as dashboards and reports, so supervisors and managers work from the same picture.

Compared with manual logging, there are three key differences:

• Time resolution. Short, frequent minor stops are rarely written down, yet together they can add up to a lot of lost time.
• Objectivity. The data comes from machine signals, not from someone's memory at the end of a long shift.
• Speed of response. Problems become visible during the shift instead of in a month-end summary.

This data is the foundation for calculating OEE you can actually trust. If the metric is new to you, see What is OEE? How to calculate it and the six big losses.

What data should you collect from your machines?

You don't need everything on day one. Focus on data that answers real production questions. Common data groups include:

• Machine state such as Running, Idle, Stop, Alarm and Setup, the core input for Availability.
• Part count and cycle time, compared with the standard cycle time to measure Performance.
• Alarm codes and messages, to find the causes behind recurring problems.
• Downtime reasons such as waiting for material, tool change, waiting for QC or breakdown. These usually have to be selected by operators on a terminal, because the machine cannot know why nobody loaded it.
• Scrap count, for the Quality rate.
• Detailed machine data such as the active program, spindle load or temperature, useful for deeper analysis and spotting wear trends.
• Energy consumption from power meters, showing energy per part and energy used while machines sit idle.

How to get data out of machines: three main approaches

1. Connect directly to the controller

Many modern CNC machines and other equipment have Ethernet ports and support communication protocols such as FANUC FOCAS, MTConnect or OPC UA. This lets you read status, part counts, alarms and programs in detail without extra sensors. Check which controller software options need to be enabled and which versions are supported.

2. Read PLC or I/O signals

Machines controlled by a PLC can be read over the network from PLC registers. For machines with no communication port, signals are commonly taken from the stack light or relays, for example green for running and red for alarm, plus a cycle-complete signal for counting parts. This covers the basics for OEE, though not details such as alarm codes.

3. Add external sensors

For very old machines, a current sensor on the motor supply can distinguish running from stopped, and a power meter can measure energy. The advantage is that you don't modify the machine's control circuit, but thresholds must be tuned for each machine, or the recorded states may be wrong.

In practice, most plants run machines of many makes and ages, so a mix of methods is normal, with all data brought into one system. This is where an factory automation integrator needs to survey the shop floor before proposing a solution.

Getting started with a machine monitoring system

1. Define the questions you want answered. For example, why a line misses its output target, or how to cut setup time on a group of machines. A clear goal tells you which data you need.
2. Pick a pilot area. Start with three to five bottleneck or problem machines rather than wiring the whole plant at once and ending up with data nobody uses.
3. Survey machines and network. Check controller models, communication ports, available signals, LAN or Wi-Fi coverage on the floor, and your IT security policy.
4. Agree on state and downtime definitions. Decide, for instance, how many minutes of stoppage count as downtime. Keep the reason-code list short enough that operators can choose correctly.
5. Install and validate. Compare system data with what you observe on the floor during the first weeks and adjust settings until they match reality.
6. Build a routine for using the data. Decide who reviews the dashboards, when, and how follow-up meetings work, then expand to more machines.

Common mistakes in machine data collection projects

• Collecting lots of data that nobody acts on. A good-looking dashboard creates no value unless decisions and fixes follow.
• Vague reason codes, or too much "other". This makes root-cause analysis impossible.
• Not involving operators. If people see the system as a tool for blame, they won't record reasons properly. Make it clear the goal is to fix processes, not people.
• Using the wrong standard cycle time. This makes Performance look better or worse than it really is.
• Ignoring network design and security. Connecting machines to the plant network calls for proper segmentation and access control.

Once the data is reliable, you can link it to other measures such as cost, delivery and energy within the QCDESM framework.

FAQ

Can older machines without a communication port be monitored?

Usually, yes. Signals can be taken from the stack light or relays, or a current sensor can be fitted to detect running and stopped states and count parts. This gives the basic data for OEE, though with less detail than a direct controller connection.

What is the difference between machine monitoring and energy monitoring?

Machine monitoring focuses on machine state, part counts and stoppages, while energy monitoring focuses on electricity use. Combining the two shows energy per part and the energy wasted while machines run idle.

Do we need to replace machines or controllers first?

In most cases, no. These systems are designed to work with existing equipment. You may need to enable a controller software option or add signal hardware, depending on what the site survey finds.

If you want to start collecting machine data, measure OEE or track energy use in your plant, the NKE team can survey your shop floor and recommend an approach that fits the machines you already have. Contact NKE through the form, call 02-746-7196 or message us on LINE @nke.th.