Building a Forensic Investigation Workbench for Industrial Vision Systems

0:00
/3:22

nvestigating Industrial AI Systems

In most industrial AI systems, the focus is on real-time monitoring.

Dashboards display live video feeds.
Alerts trigger when thresholds are exceeded.
Automation systems react to sensor signals.

But when something goes wrong, a pump behaves unexpectedly, a fill level drops suddenly, or an alarm triggers, the real question is not simply whether the system detected it.

The real question is:

What actually happened?

Answering that question often requires reconstructing events across multiple sources:

  • camera recordings
  • AI detection results
  • sensor time series
  • actuator commands
  • alarm history

To address this problem, we built Investigation Mode in the EvoLab Platform.

It is designed as a forensic workbench for industrial experiments and processes, allowing engineers and operators to analyze historical system behavior in a structured way.

Rather than presenting isolated logs or charts, the platform reconstructs a synchronized timeline of events across the entire system.


Investigation Mode Overview

The Investigation page allows users to review historical activity across one or more camera feeds within a defined time window.

Its primary objective is to transform raw operational signals into structured, explainable evidence.

Typical investigation workflows include:

  • replaying video around incidents
  • correlating alarms with sensor signals
  • analyzing actuator activity
  • capturing snapshots from recordings
  • documenting findings and observations
  • exporting a structured investigation report

These tools allow operators and engineers to move from raw telemetry to a coherent understanding of system behavior.


Why Investigation Tools Matter

Most monitoring systems focus primarily on detection.

However, detecting an anomaly is only the first step.

Understanding the cause requires:

  • correlating multiple signals
  • reviewing visual evidence
  • reconstructing system behavior over time

Investigation Mode was designed to support this process.

By combining video playback, signal correlation, event timelines, and structured findings, the platform transforms distributed telemetry into a coherent forensic narrative of the system.