Cookie Notice

We use cookies to make parts of our website work, and to improve your visitor experience.
If you only allow necessary cookies, some features of our website may not work.



Articles

Smart factories get smarter with AI inspection systems

Tanya Weaver

Stock image. Artificial intelligence can deliver sophisticated inspection systems in manufacturing, helping to improve quality and productivity (Credit: Shutterstock)
Stock image. Artificial intelligence can deliver sophisticated inspection systems in manufacturing, helping to improve quality and productivity (Credit: Shutterstock)

The smart factory is a key feature of Industry 4.0, and key to the smart factory is the ability for the things inside it to be connected in real time in what is known as the Industrial Internet of Things (IIoT).

In this intelligent environment, data flows from connected devices, machines and sensors to the network making it available to users. To process it, artificial intelligence (AI) capabilities can then be used to convert this data into actionable insights.

Cambridge-based start-up Turation is developing technology that offers a synergy between AI and IIoT technologies in what is known as AIoT. “In our AIoT system, AI takes on the burden of analysing the data. It can act on its own and much faster, so operators and decision makers see only reports with insights already extracted from the raw data,” explained Alex Kotelnikov, CEO and co-founder of Turation. 

This AIoT solution is focused on various applications including automated realtime inspection, predictive maintenance of equipment or machine control. As a result, it can help factories to improve quality and productivity, and reduce material waste and machine downtime.

Generalised solution

The key innovation of Turation’s system is its Generalised AI engine, which overcomes the limitations of today’s Narrow AI that can only perform a single task. Often based on supervised machine learning, Narrow AI’s knowledge is mostly limited to training examples. This lack of generalisation and adaptability ultimately leads to more data and training being required as the working environment evolves.  

On the other hand, Generalised AI learns concepts not labels, and so requires fewer data samples for training and, as a result, less or no human supervision. “Generalised AI is more robust and is more adaptable to new data and has lifelong learning capabilities. It can also switch between multiple tasks or domains, which means it’s much better suited to the needs of manufacturing,” said Kotelnikov.

As opposed to being in the cloud, Turation’s AIoT platform is located at the edge, which brings AI closer to the data coming from the sensors within the factory. Raw data, such as high-resolution footage from an industrial camera, does not leave the edge AI node and it is not transmitted over the network until it is specifically requested. Instead, the edge AI node sends out only data processing or inspection results of a much smaller size. 

This offers several benefits. For instance, it reduces network traffic and requirements to communication bandwidth. It offers more reliable and stable operations through local network or direct connection between sensors and AI nodes. It also presents better data security by design, which is often a basic requirement as manufacturers prefer to keep most of the data on their premises or, at least, have better control over it. Lastly, Edge AI provides a rapid response to events detected by the sensors, which can be critical for high-speed production lines and precise robotics. 

Detecting defects 

Turation has worked on a number of use cases, with the main one being automated surface defect detection and classification within the steel industry. Steel offers wide variability in terms of the type of products, their differing chemical compositions and the way they are processed. All these may result in different surface defects. 

“All this makes it challenging for mainstream AI and machine learning to perform quality inspection. Turation Generalised AI is better prepared to face such great variability,” said Kotelnikov. 

“The AI engine can also issue an alarm or even stop the production line if defect severity goes beyond a critical level. Only information of the defects detected, such as their position and type, is transmitted further by the AI node. The operator can see the total and per-line number of defective items, the distribution of quality grades based on type, size and number of defects, the distribution of defect types, the distribution of defects over time and much more, depending on customer needs.”


Want the best engineering stories delivered straight to your inbox? The Professional Engineering newsletter gives you vital updates on the most cutting-edge engineering and exciting new job opportunities. To sign up, click here.

Content published by Professional Engineering does not necessarily represent the views of the Institution of Mechanical Engineers.

Share:

Read more related articles

Professional Engineering magazine

Professional Engineering app

  • Industry features and content
  • Engineering and Institution news
  • News and features exclusive to app users

Download our Professional Engineering app

Professional Engineering newsletter

A weekly round-up of the most popular and topical stories featured on our website, so you won't miss anything

Subscribe to Professional Engineering newsletter

Opt into your industry sector newsletter

Related articles