Case Study · Modeled Scenario · Radio Degradation

Eight radios drifted for two months. Nobody noticed — until they noticed all at once.

A modeled-scenario walk-through of how a small midstream operator running 84 remote wellsites watched eight radios fail simultaneously after a quiet 9-week degradation window. This is what predictive operational intelligence would have caught — and what conventional SCADA didn't.

Aevus / Intrepid LogicModeled scenarioFor engineers · executivesPublished 2026-05-21

Modeled scenario — read this first. The operator below is composite, not an actual customer. The radio types and degradation patterns are real (drawn from our advisory board's combined field experience). The numbers are modeled from realistic operational data, not measured from a specific deployment. Aevus is pre-revenue; we don't claim a customer outcome we don't have. We claim the pattern the modeled scenario demonstrates is real.

The setup

A regional midstream operator we'll call "Trinity Energy" runs 84 remote wellsites across three Texas counties. Each site has a Schneider SCADAPack 333 RTU communicating over licensed UHF radio to a single master station near the company's gas-processing plant. The radios were installed in 2019. The master station was upgraded in 2023 to a Freewave FGR2-PE. Comms had been "rock solid for two years," per the SCADA team.

Trinity's SCADA system — built on AVEVA Wonderware — surfaces communication status per site as a boolean: green dot for "comms OK," red dot for "comms FAIL." Operators see the dots on the L1 area overview. Anything between green and red is invisible.

Remote wellsites

UHF

Licensed Part 90 (~451 MHz)

Years since last RF survey

SCADA operators on duty

What the radios were doing

Every radio in the field publishes diagnostic telemetry alongside its primary process data — RSSI (received signal strength), SNR (signal-to-noise ratio), retry count, and forward error correction (FEC) statistics. These are pollable via the same comms link the SCADA already uses. They are not displayed on Trinity's HMI. They live in the historian, unread.

Starting in early March of the modeled scenario, eight sites began showing the same pattern: RSSI drifting -1 to -2 dB per week, with no change in retry count or link status. Comms still passed. The boolean stayed green. The trend was visible only by looking at the historian numbers — which nobody was looking at.

What was actually happening

The eight affected sites shared a common feature: they all routed through a single repeater tower at the highest-elevation point on Trinity's range. That repeater's UHF antenna had developed a slow leak in the coax weatherproofing — a hairline crack in the connector boot from a hailstorm the previous September. Water was gradually corroding the antenna feed. Return loss was climbing 0.5 dB per week. Effective radiated power was dropping proportionally. Sites farther from the repeater felt it first — closer sites had margin to spare.

Timeline of the drift

WEEK 0

Hailstorm. Repeater tower antenna takes light hail. No visible damage. Hairline crack in connector boot, undetected.

WEEK 2

First measurable RSSI drop on a distant site. Site 47 (the farthest of the eight) sees its 24-hour RSSI average drift from -82 dBm to -84 dBm. Well within margin. SCADA shows green.

WEEK 5

Drift spreads. By now, six of the eventual eight sites are 4-6 dB lower than their February baselines. Still well above the radio's documented sensitivity. Still green. Still invisible to operators.

WEEK 7

First flickers. Site 47 starts seeing brief comms drops at sunset (refraction-induced multipath worsens with low signal margin). Three to five second outages, recovered automatically. The HMI shows a transient red and clears. Marked "intermittent — monitor" in the operator log.

WEEK 8

Investigation requested. The two flickering sites get added to a "look at next visit" list. No emergency. Operations is fully consumed by an unrelated compressor station overhaul.

WEEK 9

A windy Sunday. Sustained 35 mph winds rotate the repeater antenna by a fraction of a degree (the wet mounting hardware had loosened). Effective radiated power drops another 3 dB instantaneously. Eight sites go hard-offline within 22 minutes of each other.

2:47 AM MONDAY

Night-shift operator gets eight comms-failure alerts. Tries to acknowledge them individually. The pattern looks like a regional outage, not a per-site failure. Calls supervisor. Supervisor calls field tech. Field tech is two hours away.

6:14 AM

Custody-transfer measurement gap. Two of the offline sites are custody-transfer points feeding a pipeline. The measurement is in dispute for 3 hours and 27 minutes. Operations files a PHMSA-reportable communications-loss event.

11:00 AM

Field tech arrives at repeater. Diagnoses corroded antenna feed. Replaces connector and weatherproofs. Sites recover one by one as RF margin returns. Recovery complete by 1:40 PM.

The cost

Modeled, but representative of real operations of this size:

11h

Total outage duration across the 8 sites

$47K

Lost / disputed flow value in custody-transfer window

$8K

Emergency dispatch + Sunday-night labor

PHMSA reportable event (regulatory exposure)

Customer phone calls about delivery schedule

14h

Post-incident review across operations + IT

Total modeled cost: ~$70K direct, plus the PHMSA exposure (which is rarely a fine but is always a regulatory-record entry that follows the operator).

What Aevus would have surfaced

Aevus ingests the radio diagnostic telemetry alongside the process data. Three signals in the modeled timeline that the platform would have flagged:

Week 5 — fleet-wide pattern detection

By week 5, six sites had drifted in the same direction by 4-6 dB over the same 5-week window. Independent site monitoring missed this because each individual drift was small. Aevus's fleet-wide correlation engine would have flagged "six geographically-clustered sites are degrading at the same rate" as a Severity-2 advisory, recommending an inspection of the shared infrastructure (the repeater tower).

"We watch for synchronized degradation across sites because synchronized degradation is almost never coincidence. When six radios drift in lockstep, it isn't the radios — it's the path they share."— from the Aevus internal alerting rulebook

Week 7 — escalation as flickers appear

The brief sunset outages on Site 47 would have triggered an automatic escalation from Severity 2 to Severity 1: an actively-degrading repeater path with confirmed outage events. Aevus would have notified the on-call supervisor by SMS with an actionable summary and a recommended dispatch window — Tuesday or Wednesday daylight, before the next weather front predicted for the following weekend.

Week 8 — explainable recommendation, not just an alert

Aevus's alert content would have included the correlated evidence: the six-site drift pattern, the recent hailstorm weather event, the historical-tower antenna-replacement schedule (last replacement: 2019, recommended at 5 years for the model in use), and a confidence estimate. Not "something is wrong" — "the repeater tower antenna is degrading; here's why we think so; here's what to do."

The modeled outcome

With predictive intelligence on RSSI/SNR/retry, the maintenance dispatch happens on a Tuesday afternoon. The corroded connector is replaced before it fails. No 2:47 AM calls. No custody-transfer dispute. No PHMSA-reportable event. Total spend: ~$1,200 for a planned field visit. Total avoided cost: ~$70K plus the regulatory record.

Why the conventional SCADA missed it

Trinity's SCADA isn't bad — it's typical. Three specific blind spots, each common across the industry:

Boolean comms status. The HMI shows "green" or "red." There is no "yellow" for drifting. Operators with hundreds of sites can't visually scan numerical RSSI trends; they need the system to do that for them.
Per-site monitoring, not fleet-wide. Each site's drift was small. The pattern only emerges when you correlate across sites. Trinity's SCADA renders per-site dashboards; it doesn't compute fleet-wide synchronized-degradation signals.
Reactive maintenance posture. The radio team operates on a "fix on fail" budget. Predictive dispatch isn't a budgeted activity. That's not a technology gap — it's an organizational one. But the technology has to exist before the budget can shift to use it.

"We had the data. We had it for nine weeks. It was sitting in the historian. Nobody was looking at it because nobody had a reason to look at it. Aevus's pitch isn't 'we collect more data' — it's 'we tell you which slice of your existing data matters, and when.'"— Aevus founder narrative framing, on the recurring operator pattern this case study models

What this case study is, and isn't

What it is: a modeled walk-through of a real failure pattern, with modeled but representative cost figures, showing the decision-support gap that predictive operational intelligence closes.

What it isn't: a paid testimonial from a deployed customer. Aevus is pre-revenue. The first paid pilots are in active conversation as of mid-2026. When we have measured outcomes from a real deployment, we'll publish those — clearly labeled as measured, with the customer's permission, with the customer's name on the case study.

We chose to publish this modeled case study because the underlying pattern is real (verified by the Aevus advisory board's combined field experience across midstream, water, and electric utilities), and because evaluators need to see the decision-support gap clearly to evaluate whether closing it is worth a pilot conversation.

If this pattern is familiar.

If your operations team has watched a slow drift turn into a 2 AM phone call — and you want to be the operator who catches the drift, not the one who gets the call — that's the conversation we're built for.

Request a pilot conversation →Radio telemetry primer Predictive failure (companion)

Radio Telemetry

Bands, degradation, and predictive failover for field radios.

Live Telemetry Map

Every site, radio path, and degrading link in one view.

Predictive Maintenance

Watch AI catch degradation that classic SCADA misses.

See Aevus in action →

Ready to see how this applies to your operation? Start a pilot conversation — no commitment, no field changes.

← All articles in the Learning Hub