
Plenty of companies use AI to polish text or to search. Nexigna does the opposite: AI on top of your real data, meters, sensors, EMS and logs. We link those signals, spot problems and explain them in plain language. The AI prepares, the technician decides.
Nexigna links machine data, alarms, history and documents. So you don't just get an alert, but context, a likely cause and a concrete next step. The examples below run on our own geothermal cascade in Munich.
The residential complex in Munich runs on six geothermal heat pumps (2x3 cascade). Identical units, so the AI can compare them against each other. The moment one drifts off, it stands out.
Unit 3 - heat exchanger likely fouling. Approach temperature rose from 1.9 K to 4.9 K over 6 days, COP -11%. Low pressure normal, so no refrigerant indication. Units 1, 2, 5, 6 stable.
Unit 4 - compressor short-cycling. Starts per hour up to 5.8 (peers ~2.2). Check buffer or flow control.
Source stable. Source temperature and flow within baseline across all six units.
Unit 3's high-pressure cut-out tripped. A raw alert with no context.
Pulls the data window around the alarm: condenser temperature, flow, approach temperature, 6-day COP trend, compressor starts. Compares unit 3 with units 1, 2, 4, 5, 6.
Alarm triage on BMS/PCS, monthly report and an explanation of why the battery didn't charge.
Turns voice notes into a structured activity log, checks for missing fields, ensures spraying compliance.
Reads tariffs, capacity charges and volumes, compares months, flags anomalies.
Reads 3 to 5 quotes, pulls out kWh, kW, warranty, EMS, NOB and exclusions into a table.
Groups raw alarms and logs, estimates the cause and escalates only what matters.
Answers from your own documents and project folders, with source references.
We show the principle here, not the full method. How we link, model and validate the data is our engineering. We discuss that in a conversation.
Tell us what data you already have (sensors, EMS, invoices, documents). We'll show which workflow delivers value fastest.