High-Traffic Format · Technical Perspective · Published: October 29, 2026
Table of Contents
Key Takeaways
- Alarm load has a measured human limit. EEMUA 191 puts the steady-state ceiling near one alarm per operator every ten minutes, which works out to about 150 a day, and asks for no more than ten alarms in the first ten minutes of an upset [1].
- The same guide wants standing alarms held under ten per operator, with a priority split around 5% high, 15% medium and 80% low [1]. Those numbers are a sanity check you can run on your own alarm table on Friday afternoon.
- Remote SCADA access with automatic alerts by text, voicemail or email lets a small crew supervise several treatment sites without driving to them [2].
- Remote monitoring is telemetry and visibility, not closed-loop control. It tells you dissolved oxygen in basin 2 fell to 1.2 mg/L. Starting the blower is still a human decision, so decide before Friday who acts on each alert.
- Four items cause most Monday-morning surprises: a day tank that ran dry, an alarm route nobody tested end to end, a control loop left in manual since Wednesday, and a UPS that had quietly stopped holding charge.
Treatment systems fail on weekends with a regularity that stops feeling like coincidence after a few winters. Friday afternoon is when a loop gets parked in manual “just until Monday”. It is when the caustic tote that looked fine on Thursday turns out to have been the last one. It is when an alarm list that nobody has read all week finally meets an empty room.
Unattended operation is not the risk. Unprepared unattended operation is. What separates the two is a checklist that takes ten to fifteen minutes and turns the weekend from a gamble into something you are actually watching.
1. The Friday Walk-Through: Ten Checks Before You Leave
| # | Check item | Where | Pass criterion | If it fails |
|---|---|---|---|---|
| 1 | Chemical tank levels | Day tanks, tote levels | Enough stock for weekend consumption plus one full day | Arrange delivery or transfer; confirm low-level alarm setpoint above expected weekend draw |
| 2 | Dosing pump operation | Each dosing skid | Stroke/flow matches dosing setpoint; no airlock | Switch to duty/standby pump; verify anti-siphon valve |
| 3 | Alarm routing | SCADA/PLC alarm table | Every enabled alarm routes to a phone that will be answered this weekend | Update contact list; test one live alert end-to-end |
| 4 | Remote access & data backfill | SCADA server / cloud | Live values current (<5 min old); trend data uploading | Restart gateway; confirm cellular or VPN link |
| 5 | Manual/bypass mode audit | HMI mode summary; field switches | Zero loops left in manual or bypass from the week’s interventions | Return to auto; document any intentional bypass with expiry |
| 6 | Duty/standby equipment status | Pump and blower panels | Standby units available, auto-changeover enabled, no tripped breakers | Reset or lock out; align spare parts |
| 7 | Softener/regeneration status | Control valve heads | Sufficient brine and reserve capacity for weekend demand; regeneration trigger tested | Trigger a manual regeneration before leaving |
| 8 | Flood & leak defenses | Sumps, float switches, drain valves | Sump pumps on auto; high-level float functional; no hoses or drains left open | Clear; verify float switch by lifting it |
| 9 | Power resilience | UPS, generator, ATS | UPS charged; generator fuel and auto-start armed | Refuel or defer; note which loads ride through |
| 10 | Last-10-minute patrol | Full site walk | Screens match field state: no unusual noise, smell, or weeping joints | Fix or escalate before the doors lock |
The table is the checklist. The rest of this article explains the four items that fail most often, and how to make each one fail loudly instead of quietly.
2. Chemical Stocks and Levels: The Weekend Math Nobody Does
Chemical starvation is the most common self-inflicted weekend outage, and the arithmetic is dull enough that people skip it. Friday 16:00 stock, minus weekend consumption, minus Monday morning consumption before the delivery truck shows up. That is the whole calculation.
Chlorine, antiscalant, caustic and defoamer run down at different rates, because dosing is usually paced to flow and weekend flow is lower but almost never zero. A plant running at 20% of design flow still empties a tote, just later than you expected.
Two safeguards turn this from a mental exercise into something mechanical. First, set low-level alarms above the expected weekend draw, so the alert fires while there is still time to act, not when the pump is already pulling air. Second, trend dose against level. If the tank drains faster than the flow-proportional prediction, you have a passing valve or an over-stroking pump, and Friday afternoon is the worst possible moment to discover it.
Plants that dose against measured water quality rather than a fixed schedule pick up a second benefit here. When the dose follows the load, the tank draw follows it too, which widens the gap between “low” and “empty” without anyone adding a margin on top of last month’s margin.
3. Alarm Routing: The Question Is Not Whether Alarms Exist, but Who Hears Them
Every treatment system has alarms. Far fewer have an answer behind them.
Alarm management has spent thirty years quantifying how many notifications a person can actually absorb, and the numbers are lower than most plants assume. EEMUA 191, the guide most control rooms are measured against, treats about one alarm per operator every ten minutes as the steady-state target, with a peak of ten in the first ten minutes of an upset [1]. ANSI/ISA-18.2 and its international twin IEC 62682 go at the same problem from the design side: an alarm is a notification that requires an operator response, and anything that does not require a response is an event, not an alarm [1].
A weekend water system that sends forty nuisance texts between Friday night and Sunday morning has trained whoever holds that phone to ignore all forty, including the one that matters. So Friday’s check is an exercise in triage:
- Every enabled weekend alarm needs an action and an owner. If nobody can name what they would do when it arrives, disable it or rework it. An alarm without a defined response is not protection, it is noise.
- Test the chain, not the configuration. Push one live test alert through the entire route, sensor to PLC to SCADA to SMS to phone, and get a human to acknowledge it. A distribution list is not a response plan until someone replies.
- Check the escalation timer. If the first phone does not acknowledge within ten or fifteen minutes, the alert should climb to a second contact automatically. Most telemetry platforms support this natively. It is one checkbox, and it decides whether a Saturday overflow is a half-hour fix or a Monday investigation.
4. Remote Access and Data Backfill: Watch It, Don’t Just Visit It
Weekend instrumentation exists so the system can describe itself while nobody is standing next to it. That needs three capabilities verified, not assumed.
Current values. The gateway has to be online and the newest reading has to be fresh. Stale data is worse than no data, because it looks like surveillance while actually reporting history.
Backfilled trends. If the network drops for four hours on Saturday, does the logger hold the gap and re-upload it, or is it gone? Most modern IIoT platforms buffer and backfill. Confirm that yours does.
Parameter coverage. Level, flow, pressure, and the chemistry that defines compliance (pH, conductivity, dissolved oxygen, residual chlorine, turbidity) should arrive as one synchronized stream rather than a patchwork assembled from six devices. Multi-parameter probes reporting over industrial protocols, Modbus on the plant side and MQTT to platforms such as AWS IoT Core or Azure IoT Hub, were built for exactly this: one probe, one stream, every parameter correlated in time. See the practical benefits of IoT-enabled water quality sensors. The same synchronization is what makes the data useful on Tuesday as well as Sunday.
Be clear about what remote monitoring does not do. It is telemetry and visibility, not closed-loop control. It reports a dissolved oxygen drop; acting on that reading is a separate, human step. That division of labour is fine, provided Friday’s check confirmed there is a human on the other end of it.
5. Bypasses, Manual Modes, and the Duty Pump You Forgot
Friday afternoon modifications are the quiet saboteurs. Every loop left in manual since Wednesday’s maintenance, every isolation valve cracked for a filter change, every hand switch that was supposed to be temporary removes a layer of automatic protection for the next sixty hours.
The audit is mechanical. Call up the mode summary on the HMI, walk the field switches, and return every loop to automatic unless the bypass is documented, justified and time-stamped with a return-to-service owner.
The same walk covers duty/standby equipment. Standby pumps exist for weekends. Confirm each unit is available, not tripped and not locked out for maintenance. Confirm automatic changeover is enabled. Confirm the pressure or flow switches on the standby side actually prove rotation, because a standby pump that failed silently in March is not a standby pump. If the train includes softening or filtration heads, check brine and reserve capacity for weekend demand and trigger a manual regeneration before you leave if the controller is due.
6. Water Where You Don’t Want It, Power When You Need It
Flood and power failures share a property: both turn a small local fault into a multi-day outage, and both are cheap to prepare for and expensive to improvise.
Flood defenses. Sump pumps on automatic, high-level floats physically tested (lift the float, watch the pump start), hose stations and drain valves closed, no equipment running unattended over an open pit. A hose left in a tank on Friday is a siphon by Saturday night.
Power resilience. Confirm the UPS batteries hold charge, the generator has fuel and its auto-start is armed, and the automatic transfer switch has actually transferred in test mode at some point this year. Decide deliberately which loads ride through an outage and which do not: dosing and telemetry usually should, comfort loads usually should not. If the site loses grid power on Saturday, the alarm chain itself has to run on something, which makes the UPS the first thing to check and the last thing to economize on.
7. The Last Ten Minutes
The final patrol is a habit, not a formality, and it follows a fixed route so omissions become visible: chemical storage, dosing skids, filter and softener area, pump room, switchboard, sumps, control room. At each stop one question: does the field match the screen? Screens show what the system believes. The walk shows what is true. Unusual noise, a weeping joint, a hot motor, chlorine where there should be none, any one of these is worth ten minutes on Friday.
Then lock the doors, open the alarm summary one last time, and check the standing-alarm count. If the screen already shows a dozen acknowledged-but-active alerts, the weekend has started badly and nobody has even left the building. Clear them or re-prioritize them before the room goes dark.
References
- EEMUA Publication 191, Alarm Systems: A Guide to Design, Management and Procurement (4th edition, 2024), and ANSI/ISA-18.2-2016 / IEC 62682, Management of Alarm Systems for the Process Industries: steady-state alarm rate near 1 per operator per 10 minutes (about 150 per day), peak of 10 alarms in 10 minutes during an upset, fewer than 10 standing alarms per operator, and a priority distribution around 5% high / 15% medium / 80% low.
- High Tide Technologies — “SCADA Solutions for Water & Wastewater Treatment Plants”: automatic alerts by text, voicemail or email and remote access to live plant data for treatment plants. https://htt.io/resources/water-treatment-plant-basics/
- Shanghai ChiMay — practical benefits of IoT-enabled water quality sensors: multi-parameter probes, Modbus and MQTT data paths.
About the author: Written by the Shanghai ChiMay Technical Editorial Team — engineers who support unattended operation of water treatment trains with multi-parameter water quality instrumentation, softening and filtration control valves, and remote-ready digital outputs.
