Putting an Online Gauge on the Line: Integration, Data and Sorting

Buying an online gauge is the smaller half of the job. The larger half is everything between the measurement and the decision it feeds: when the gauge takes a reading, where the number goes, what happens to the part that failed, and who sees the trend. This page covers the integration questions that decide whether an inline installation produces control or just produces data. It deliberately does not quote interface or protocol specifications — those are model- and configuration-specific, and the honest instruction is to confirm them with the vendor for the exact unit you are buying.

Start From the Decision, Not the Instrument

Before specifying anything, write down what happens to a reading. Three patterns cover most installations:

• Monitor and trend. Every part is measured, the data is logged, and action is taken when the trend moves. No automatic rejection. This is the usual first step and the cheapest to commission.
• Alarm on limit. Out-of-tolerance readings raise an alarm for an operator. Needs defined limits and a defined response, and a plan for alarm fatigue.
• Automatic sort. Readings drive a physical reject. This is the most demanding: it needs a known and bounded delay between the measurement point and the reject point, and a reject mechanism fast enough to act within it.

The Delay Budget

Automatic sorting fails more often on timing than on measurement. The part is measured at one point on the line and rejected at another, and the distance between them, divided by line speed, is the time you have to make and execute the decision. Add to that the measurement and processing time itself and the response time of the reject mechanism. If the total exceeds the available travel time, the wrong part gets rejected. Work this out during design — it constrains where the gauge can physically mount, and it is not something a faster instrument fixes.

Triggering and Sampling Discipline

A gauge running at its published rate — 2400 samples per second on the LM-1030 — produces far more data than a control system wants. Decide deliberately whether you are storing every sample, storing statistics over a window, or storing exceptions. Storing everything is rarely useful and always expensive to keep. Storing only pass/fail loses the trend that would have warned you before the failure. The pattern that usually works is statistics at a fixed interval plus exception records, with the raw stream available for a limited retention period when you are diagnosing something.

Data Output: What to Confirm Before You Buy

The following are configuration-specific and must be confirmed for the exact model and controller you are ordering. They are listed as questions to put to the vendor, not as specifications of any POMEAS product:

• Which physical interfaces and protocols the controller supports, and whether the one your line uses is among them;
• Whether analogue output, digital I/O, serial or industrial Ethernet is available, and in what combination;
• The maximum update rate at which measurement values can be delivered to an external system — which may be lower than the internal sampling rate;
• Whether the controller can buffer data during a network interruption, and what happens to readings taken during one;
• What the software exports, in what formats, and whether SPC export is included or licensed separately.

Treating any of these as assumed rather than confirmed is the most common cause of an integration that overruns.

Environment and Mounting

The LM-1030 head carries an IP64 enclosure rating — protected against dust ingress and splashing water. That suits most factory environments but is not a washdown or immersion rating, so a line that is hosed down needs a different arrangement or additional protection. Mounting has to hold the transmitter-to-receiver distance at 160 ± 40 mm and keep the product inside the beam. Vibration from neighbouring equipment transfers straight into the reading, so the mounting structure is a measurement component, not a convenience.

Keeping It Trustworthy After Commissioning

The throughput question sits alongside this one: how many gauging stations a target rate needs is worked through in cycle time and station count, and where the measurement should physically sit is argued in inline versus offline optical measurement. Commissioning proves the installation works once. Staying trustworthy requires a verification interval against a certified artefact, a cleaning interval for the optical windows, and a defined response when verification fails. Those belong in the maintenance plan before handover, not after the first bad batch. See calibration and measurement stability.

FAQ

What interfaces does the gauge support?

Confirm this with the vendor for the exact model and controller configuration you are ordering. Interface and protocol support is configuration-specific, and assuming it is the most common cause of integration overrun. This page lists it as a question to ask rather than a specification to quote.

Can the gauge drive a reject mechanism directly?

That depends on the controller's digital I/O and on whether the timing budget works out. Check the delay between the measurement point and the reject point against the line speed before assuming it can.

How much data should we store?

Usually statistics over a fixed interval plus exception records, with the raw stream retained briefly for diagnosis. Storing every sample is expensive and rarely used; storing only pass/fail discards the trend that gives early warning.

Is the 2400 samples per second rate the rate we get on the network?

Not necessarily. The published rate is the internal sampling rate; the rate at which values can be delivered to an external system may be lower. Confirm the external update rate specifically.

Does IP64 mean we can wash the area down?

No. IP64 covers dust ingress and splashing water. Washdown or immersion needs a higher rating or additional protection.

What should be in the maintenance plan?

A verification interval against a certified artefact, a cleaning interval for the optical windows, a defined response when verification fails, and periodic checks of mounting stability and pass-line position.

Next Step

Send the line layout, line speed, what should happen to a failing part, and which control system the data has to reach. We can confirm what the measurement side can deliver and flag the integration questions that need answering before an order. Begin with the method selection guide, or the LM-1030 product page for the diameter-gauge route.