IKA — Platform Overview

IKA (Industrial Knowledge Automation) is a Retrieval-Augmented Generation (RAG) AI system that classifies industrial signals and assets for power generation plants — hydro, wind, solar, storage, and hybrid — according to two standards:

• IEC 61850 — the international standard for substation/plant automation, which defines Logical Nodes, Data Objects, Data Attributes and Common Data Classes used to describe what a signal is informatically.
• IEC 81346 — the standard for structuring a plant's functional hierarchy (Main System, Technical System, Component System), used to describe where an asset sits in the plant.

Every signal the platform classifies is reduced to a single, dense 8-field code:

main_sys . tech_sys . comp_sys / lnclass lninstance . do . da . cdc

Field	Standard	Meaning
main_sys	IEC 81346	Main functional system, e.g. a generation unit (M01) or a hydraulic work (L01). Always mandatory.
tech_sys	IEC 81346	Technical subsystem within BL0, e.g. the turbine (MEA00) or the governor (MEX00). - if not applicable.
comp_sys	IEC 81346	Component / measurement type within BL1, e.g. a temperature reading (BT001). - if BL1 is also -.
lnclass	IEC 61850	Logical Node class, e.g. MMXU (measurement), CALH (alarm), XCBR (breaker control).
lninstance	IEC 61850	Numeric instance of that Logical Node (a per-plant routing/client-level convention, not a unit number).
do	IEC 61850	Data Object, e.g. TmpSv, Str, OpAlm.
da	IEC 61850	Data Attribute, e.g. mag.f, stVal.
cdc	IEC 61850	Common Data Class that types the (do, da) pair, e.g. MV, SPS, ALM.

Worked example: a description like "Temperatura cuscinetto turbina unità 2" (turbine bearing temperature, unit 2) resolves to something like M02.MED00.BT001/STMP1.TmpSv.mag.f.MV — unit 2's main system, the turbine bearing technical system, a temperature analogue component, read through an STMP measurement Logical Node.

What you can do in this app

The sidebar is organised in two groups:

• Classify — day-to-day classification work:
  • Classify — classify one signal description at a time, with full reasoning and confidence detail. See Classification Modes.
  • Classify Bulk — upload a CSV of signals and classify them all, either hands-off or with operator review of uncertain rows. See Classification Modes.
• Config — knowledge, plant data, and pipeline configuration that make classification accurate over time:
  • Docs — this section.
  • Knowledge Base — the normative and utility documentation that grounds classification. See Knowledge Base & Documentation.
  • Vocabulary — domain term-expansion lists used during pre-processing. See Context Enrichment & Plant Memory.
  • Plants — upload and review plant schematics (P&ID, single-line diagrams). See Plant Schematics Ingestion.
  • Plant Collections — browse and manage each plant's operator-validated signal history. See Context Enrichment & Plant Memory.
  • Reviewed Signals — the cross-session log of every signal an operator has corrected or confirmed. See Learning from Corrections.
  • Vector DB Config — sync the underlying reference/master/validated data into the vector database. See Context Enrichment & Plant Memory.

In short: Classify is where you get results, Config is where the system's knowledge — normative, plant-specific, and historical — lives and grows.