AI software for early detection and diagnosis of process deviations in the process industry

KIIO

Lausanne (CH)New on Amplifa

EPFL spin-offOrigin

Lausanne, SwitzerlandLocation

About KIIO

KIIO is an EPFL spin-off based in Lausanne (Switzerland) that develops interpretable AI software for proactive process reliability in the chemical industry and beyond. The technology is based on physics-informed graph AI, which uses a plant's P&ID (piping and instrumentation diagram) as its physical structure and learns the normal process dynamics from operational data. The software detects emerging process deviations early, identifies whether a fault originates in a sensor, actuator or the process itself, quantifies severity and diagnostic confidence, and provides process and maintenance teams with recommended actions. It connects read-only to existing process data without additional hardware; the model learns self-supervised from normal operation without historical fault examples. The founding year is not specified.

Site: Lausanne
Processes: AI-based early detection of process deviations, Diagnosis of fault causes (sensor, actuator, process), Quantification of severity and diagnostic confidence, Recommended actions for process and maintenance teams, Offline proof-of-value studies, Connection to existing process data (read-only, without hardware)
Materials: Physics-informed Graph AI, Control-aware Dynamics, Self-supervised Learning, Offline Proof-of-Value Study
Industries: Chemical industry, Process industry

01
Connection to existing process data (read-only, without hardware)
02
Detection of emerging process deviations (Detect)
03
Diagnosis of the fault mechanism: sensor, actuator or process (Diagnose)
04
Quantification of severity and diagnostic confidence (Quantify)
05
Recommended actions for process and maintenance teams (Guide)

Technology and Approach

Graph-KI

Physics-informed Graph AI

Uses the P&ID as a structural inductive bias and trains a graph neural network to learn the normal interactions of process variables from operational data.

CAD

Control-aware Dynamics

Explicitly distinguishes process measurements from actuator and control signals in order to separate control reactions from genuine process deviations.

SSL

Self-supervised Learning

Learns self-supervised from normal operation and requires no historical fault examples or predefined fault libraries.

PoV

Offline Proof-of-Value Study

Preliminary study to evaluate the user scenario before integration into the production environment.

Locations

Lausanne

HQ / EPFL spin-off

Functions

**Early detection of process deviations**Graph AI · Process industry

**Diagnosis of sensor, actuator and process faults**Control-aware Dynamics · Chemical industry

**Virtual Sensing**Physics-informed Graph AI · Process industry

**Data imputation for incomplete industrial data**Self-supervised Learning · Process industry

Industries

Chemical industryProcess industry

Additional services

AI-based early detection of process deviationsDiagnosis of fault causes (sensor, actuator, process)Quantification of severity and diagnostic confidenceRecommended actions for process and maintenance teamsOffline proof-of-value studiesConnection to existing process data (read-only, without hardware)

History

KIIO is an EPFL spin-off based in Lausanne; founding year not specified