Industry News

Adopting Artificial Intelligence in Manufacturing May Seem Challenging – But It’s Not!

Despite its huge potential, industrial AI is often considered too tricky to adopt and scale. But the reality belies expectations.

The time for industrial artificial intelligence has finally come. The world’s largest manufacturing companies are rushing to adopt AI to increase throughput, assure product quality, optimize supply chains, and improve worker safety.

But what about midsize and small manufacturers? Can they compete with the likes of Apple, Boeing, and Toyota? 

The answer is a definitive yes.

In this article, I will explore how an ordinary manufacturing company can approach industrial AI. I will look into major challenges, recommend simple AI use cases to get started, and demonstrate how one of our manufacturing clients has successfully kicked off a full-scale AI transformation.  

Three Major Barriers to AI Adoption in Manufacturing

Though AI has been around for decades, it is not a plug-and-play technology. The development, adoption, and scaling of any AI use case requires concerted effort, because:

  1. Relevant AI use cases are hard to find. Businesses do not usually know how to identify AI opportunities: they start with unclear objectives, pay no attention to data, and expect too much, too soon. Actual business outcomes and data are key to choosing the right AI use cases.
  2. Implementing AI requires specialized expertise. Hiring a team of AI professionals who can take your AI solution from concept to production is a significant investment. Thankfully, various AI services that are available in the cloud, as well as AI consultancies, can help reduce this burden.
  3. AI use cases require large amounts of data. AI algorithms learn on vast amounts of data to mimic human intelligence. Data should be accessible, usable, relevant, and unbiased. This applies to both structured (data in tables) and unstructured (images, videos, sensor inputs) data.

There are many more AI implementation challenges. For example, you may have to augment complex machinery with IoT devices, overhaul tried-and-true operations, and build a robust infrastructure for your data and algorithms.

Industrial AI Use Cases for a Quick Start

However, not all AI solutions are that complicated. Here are six AI use cases that do not require next-level technology, and can be delivered in as little as a month or two:

  1. Machine optimization. A great use case for eliminating production bottlenecks. AI and computer vision (CV) are applied to monitor machines in real-time video feeds, to detect their run vs idle time. Based on the insights, you can optimize operations (e.g. deliver materials at a more frequent rate, train your workforce) to improve machine utilization. 
  2. Predictive maintenance. A typical IoT use case can also be enabled with CV. Instead of sensor data, algorithms process video streams from cameras to spot, for example, the early signs of wear on conveyor belts or rust on idlers. Certain conditions should be met to ensure the accuracy of such algorithms. Namely, the production floor should be well-lighted; the cameras should be high-resolution and high frame rate, to accurately capture every detail in the picture.
  3. Quality control. CV can reinvent quality control across the production line. Instead of relying on manual inspection and expensive IoT, let AI algorithms learn from new video data captured by cameras, to consistently catch more defects while reducing the number of false positives.
  4. Worker safety. Creating a culture of safety and carrying out regular PPE training are not enough to protect workers against hazards. You have to enforce your safety policy. CV-powered cameras can help you monitor employees’ PPE use and adherence to safety recommendations.
  5. Supply chain tracking and visibility. With CV, you can monitor how your warehouse and production floor work in real-time. You can analyze how forklifts deliver materials to assembly lines, to identify inefficiencies like poor facility layout, inventory commotion, or disjointed operations.
  6. Product and packaging integrity. Just as with production defects, the irregularities in a product or its packaging can be hard to detect. In most cases, such integrity control operations require human review. With CV, most checkups can be automated, with only disputed cases needing review by QC professionals. 

What these AI use cases have in common is that they are enabled by computer vision; they involve cameras that capture video streams and rely on algorithms that do the thinking in the cloud. Let’s take a look at how one such CV use case can be implemented in a real-world industrial environment.

A Simple Path to AI

AI transformations can seem challenging, but sometimes you need only cameras to embark on a successful AI journey. 

AI can be simple enough to enable you to optimize manufacturing operations, step-by-step, without forcing you to rework your entire infrastructure from scratch. 

Figure out what you want to achieve with AI, check if you have the necessary data, find the right team for the job, and start to implement and scale successful AI use cases one by one across your company.


Rinat Akhmetov, Product Lead, AI worker safety

Rinat Akhmetov is the ML Solution Architect at Provectus. With a solid practical background in Machine Learning (especially in Computer Vision), Rinat is a nerd, data enthusiast, software engineer, and workaholic whose second biggest passion is programming. At Provectus, Rinat is in charge of the discovery and proof of concept phases and leads the execution of complex AI projects.