AI Failure Detection¶

On this page you will learn about Kelvin AI Failure Detection in Kelvin Core and how it fits into Kelvin Collaborative Control Software.

What is Kelvin AI Failure Detection ?¶

Kelvin AI Failure Detection is the AI / Machine Learning section. Kelvin.ai uses advance machine learning techniques and decades worth of industrial data to create advanced failure detection models for all types of machines.

Failure detection uses a semi-supervised machine learning anomaly detection system. Machine learning models have been created and tuned using many decades worth of industrial data on many types of machinery and components.

How is this Data Used ?¶

Failure detection is used in a number of areas of the Kelvin Core to provide insights into the data and its effect on your operations and processes;

Preventative Alarms¶

When a failure detection is detected, this does not mean that the asset has yet failed. In most cases, unless there is a catastrophic failure or disassembly, this alarm is a pre-warning on a potential failure of your asset.

After analyzing the alarm and its data, you can then plan proper maintenance schedule to replace or maintain the problem.

Carbon Maps¶

Carbon Maps is a built-in analytical tool that helps identify where assets are performing poorly. With this data it can calculate the monetary losses incurred by these assets and help you plan your maintenance expenditure and production efficiency against all your asset real-time performances.

It also calculates the additional carbon emissions produced by these inefficient assets through the higher electricity consumption and converts these figures into the additional carbon dioxide emissions entering the atmosphere.

This is a good tool to promote your company's ambitions to protecting the environment and limiting the carbon emissions being released.

Failure Detection Alarms¶

The Failure Detection Alarms are created by the Kelvin Machine Learning Model algorithms.

Data is not enough to create good models. The Kelvin Data Scientists also worked closely with Industrial Experts to create the right feature engineering processing on the data based on their domain experience so that failure detections are reliable.

What is Detected ?¶

Kelvin AI Failure Detection can detect failures in many different areas of a machine through direct or indirect methodology.

The Kelvin AI Failure Detection will use the available sensors in an asset to try and detect any preventative maintenance failures, imminent failures or actual physical failures on many of the parts in an asset.

Kelvin Part Detection¶

Using the Kelvin Sensors, the Kelvin AI Failure Detection will monitor the health of many of the Kelvin Parts in the Kelvin Asset.

When an anomaly is detected, you will receive a Kelvin Alarm that will identify the Kelvin Part which has the detected problem along with a description of the type of problem and possible solutions.

Kelvin Component Detection¶

If the Kelvin AI Failure Detection detects an anomaly but can not identify which Kelvin Part is causing the anomaly, you will receive a Kelvin Alarm which will identify the Kelvin Component which has the problem.

You will need to inspect the asset and identify yourself which part in the Kelvin Component is causing the anomaly.

Plug and Play¶

The Kelvin Core takes out all the hard work of applying AI to your data.

Activating the AI is as simple as a click of a button and setting a few simple a parameters. The analysis will then immediately start monitoring and analyzing your real time data stream.

Activating Failure Detection¶

The failure detection models are built into the Kelvin Assets, Components, Parts and Sensors in the Asset Templates. When you drag and drop any Kelvin Template you will have the failure detection options available in properties.

Failure detection features are available for the complete Kelvin Assets or on a per Kelvin Component level and requires minimal setup to start detecting failures.

Details of Failure Detection¶

When a failure is detected, then an Alarm will be generated. You can go to the Asset to see the details of the data which created the Alarm.

Rules for Setup Failure Detection¶

The Failure Detection has a number of rules that you need to follow while setting up (by adding New Kelvin Assets). If they are configured wrong, then you will find the Failure Detection does not work as expected.

Principles of Failure Detection¶

The Kelvin AI Failure Detection will detect failures in Kelvin Parts by monitoring any Kelvin Sensors attached to the machine.

The Kelvin Sensors do not need to be attached to the direct parent Kelvin Component where the Kelvin Part is being monitored.

For example, the bearings (Kelvin Part) on a pump (Kelvin Component 1) can be monitored by vibration sensors mounted on the AC Motor (Kelvin Component 2).

Run Validation¶

When you add a new Kelvin Asset (or add a new Kelvin Component, Kelvin Part or Kelvin Sensor to an existing Kelvin Asset), then you need to Run Validation before you can save it.

Run Validation will check that the failure detection structure and all properties have been properly setup.

List of AI Failure Detections Available¶

Centrifugal Pump¶

Failure Types¶

Required Information¶

Vibration Sensor¶

For the vibration or accelerometer sensors, you need to select which axis is in the same plane as the pump shaft. This ensures the failure detection will be monitoring the correct plane for anomalies.

Failure Detection Definitions¶

Failure Class Properties¶

Kelvin Assets¶

Kelvin Components¶

Kelvin Parts¶

Kelvin Sensors¶

Other Related Links ¶

Interesting links related to Kelvin AI Failure

• Kelvin Overview -> Overview
• Documentation -> Kelvin Core -> Kelvin Manager UI -> Management -> Assets -> Assets

Last Modified¶

Kelvin Documentation AI Support

Hi. My name is KevDocBot. How can I help you?

Send