Jump to content

Safety AI User Guide/System Overview

From Event Gates Wiki

System Overview

Chapter Overview

Learn about the main components of a Safety AI system and how they work together.

What is a Safety AI Deployment?

A Safety AI deployment is an AI system that continuously monitors one or more areas using industrial cameras and artificial intelligence. It detects predefined safety conditions and generates events when configured criteria are met.

Depending on the application, the system may also activate visual or audible alarms, relay outputs, or other connected devices.

Main Components

A typical Safety AI system consists of several hardware and software components.

Hardware components of the system:

  • Visbox – The industrial AI computer that runs the AI models and processes camera streams.
  • IPIO – Provides digital inputs and outputs for communication with external equipment.
  • Cameras – Capture live video from the monitored area.
  • Warning Devices – Optional beacons, sirens, or other devices that alert personnel when an event occurs.
  • Ethernet switches - both PoE and non-PoE switches for network communication.
  • Electrical Panel - an industrial housing for system components.

Software components of the system:

  • VIS – The user interface used to monitor cameras, review events, and manage the system.
  • Visor is the main user interface of the Safety AI for monitoring, logs and camera management.
  • Config Manager (CM) is the integrated configuration tool of the system, accessible through Visor interface.
  • VIS Admin is the integrated administrative tool for VIS again accessible through Visor.
  • IPIO Manager - web based management tool for IPIO devices. (see the new IPIO User Guide)

How the System Works

The Safety AI system operates continuously using the following process:

  1. Cameras capture live images.
  2. The Visbox receives and analyzes the camera streams using AI models.
  3. When a configured condition is detected, the system creates an event.
  4. Depending on the configuration, equipment's control units will be notified, and visual or audible warnings and output signals may be activated.
  5. Users can review the event in VIS and watch log videos.

Note


Safety AI assists operators by providing timely information. It does not replace established safety procedures or the responsibilities of machine operators.

Typical Workflows

System Workflow

Normal cycle of a regular Safety AI deployments relies on constant camera feed, processed by VIS and decided output to be set to the equpment's controller. VIS can also manipulate sound and light alarms to draw user attention. Normal workflow does not require any operator involvement and happens fully automatically.

Notification Flow

In order selected users to be notified of events without a need to access the system, Safety AI system is capable to send out event notification emails with event log videos attached. This feature requires and active internet connection.

Operator Workflow

Regular Safety AI deployments do not require operator involvement but end users can access the system via Visor application to monitor live feed or inspect event and system logs including logged event videos.

An end user typically performs the following tasks:

  1. Monitor live camera feed.
  2. Observe system notifications and events.
  3. Review event details when required.
  4. Respond according to the site's operating procedures.
  5. Continue monitoring system status.

Note

Safety AI outputs rely on the current state of the event and change without external stimulation. In case an event notification to an equipment's control unit results an equipment to stop, it is strongly suggest that a user should reset the notification before restarting the equipment.

How to Access a Safety AI System

Although every Safety AI installation may have it's own characteristics, Safety AI in general is a stand alone system with a closed local area network. Safety AI does not require corporate network or internet access. It's local network hosts it's internal components eg cameras, Visbox and one or multiple IPIOs. To access such a closed loop system users need to physically connect to the network via it's switches.

IP Settings

All nodes in a Safety AI network have static IPs. Safety AI networks do not host any DHCP servers and services which means any user who wishes to connect to a Safety AI network must set the IP address of the PC manually. Consult with system administrator for the correct IP settings.

Default

Default IP settings in Safety AI networks follow the 192.168.1.x pattern, where Visbox sits on 192.168.90 and IPIO on 192.168.1.10 (and above if more than one IPIO is available). This is not a requirement and every installation may have chosen a different IP setting therefore consult with your system admin, integrator or deployer before attempting to connect.

Changing IP address and joining a network

If you experience difficulties changing your IP address or join a WiFi network, your corporate security settings may be blocking you to do so. In that case we highly recommend you to consult your IT team.

Crane Safety AI

Crane Safety AI panel are expected to be mounted on the trolley or optionally bridge of the crane which prevents physical access to the network while on the floor. To allow access without climbing up to the crane, Crane Safety AI deployments are equipped with an access point or remote access device depending on the installation and available support contract.

Deployments with access point (AP) acts as a Wi-Fi Hotspot advertising its network SSID which requires additional credentials to join the network. Since those APs also distribute IP addresses, no static IP settings are required for this scenario.

Remote Access

Depending on the support contract your Safety AI deployment may be equipped with a remote access device which is only accessible through a secured VPN. Consult your systems integrator or deployer for remote access information.