Welcome to this installment of the IBM cheat sheet series: your jargon-busting, straight-talking guide to IoT technologies. Today, we shine a little light on the concept of Intelligent Visual Inspection, and, in particular, the IBM Visual Insights solution.
Impress your boss! Intelligent Visual Inspection in ten seconds
Before we delve into the whys and wherefores of Intelligent Visual inspection, here’s an official-sounding definition for you to memorize. Bonus points if you can say it in one breath:
‘Intelligent Visual Inspection is the process of using cognitive capabilities to review and analyze components, products and parts for defects. It works by comparing live assembly line images with previously analyzed and classified pictures of image defects, to identify possible faults. This is particularly valuable when these faults aren’t visible to the naked eye. An intelligent visual inspection tool like the IBM Visual Insights solution can reduce inspection times for manufacturers by up to 80 percent.’
Why do we need it?
Once upon a time, manufacturers relied on manual inspection to ensure their products (and parts of products) were functioning correctly, before packaging them up and sending them to market. A human would have to laboriously inspect every component for scratches, missing or misplaced parts, colour errors and other flaws. As you’ve probably anticipated already, there are a few problems with this approach. It is:
Intelligent Visual Inspection was designed to tackle these difficulties. It aids factory inspectors by automating part of the inspection process and augmenting the capabilities of the inspector.
How does it work?
Intelligent Visual Inspection comprises two main tools: images captured using high definition cameras, and IBM Watson’s AI-powered capabilities. Images of normal and abnormal products are fed into Watson’s central learning service to teach it the difference between faulty and correctly functioning components. Watson uses these data to build an analytics library of known defects that can identify quality issues and learn continuously from feedback. This technology can be used to spot defects in live images from the assembly line.
A camera captures images of product components as they move through production and assembly, and feeds these to Watson for analysis. By comparing them with the images it has already processed, Watson can detect defects as minor as tiny scratches and punctures as small as pinholes. It also knows whether parts are in the correct location and are the right shape and colour.
Watson’s classification of potential quality issues is automatically sent to manual inspectors, who perform a secondary check and can override the solution’s decision if need be.
It ain’t perfect, but it’s always self-improving
The beauty of a solution like IBM Visual Insights is that it’s always learning and improving. When manual inspectors review the tool’s classifications for errors, it takes on their feedback and plugs it in to the next training cycle for that particular analytics model. In short, this is a truly cognitive tool that learns from its mistakes, just like a human does.
With enough data, the solution becomes more accurate, which further augments the capabilities of inspectors. Training time is reduced, inspector turnover becomes less problematic, and overall accuracy increases. Not bad. And because of this self-learning capability, even very small defects can be spotted and reported automatically, saving manufacturers time and money.
It also provides cognitive insights for more efficient problem resolution in the future. This means tackling the root causes of errors to ensure the same conditions aren’t repeated.
Use case for Intelligent Visual Inspection
IBM Visual Insights is already bringing benefits to many of its early-adopting clients. An electronics company, for example, is using IBM cognitive technology to automate product inspections to boost production quality and throughput. By speeding up the visual inspection process, Watson IoT technology is helping to restore agility in the production line and remove the bottleneck caused by time-consuming inspections.
A period of just three months saw an inspection lead time reduction of over 85 percent. The combination of AI and human expertise led to improved product quality and faster, more accurate defect detection.
There are plenty of other use cases for this technology, too. The video below demonstrates how IBM Visual Insights works:
Summing up: the advantages
So there you have it: Intelligent Visual Inspection de-mystified. We’ll leave you with a summary of its three main advantages:
Inspection times reduced by up to 80%
Defects reduced by 7 – 10%
Augments the capabilities of human staff and frees them for more specialized work
How can you improve quality, prevent downtime, and increase throughput in your manufacturing environment? The obvious answer. You bring in a robot (meet ARMonk, ARMie for short), the Internet of Things (IoT), and artificial intelligence (AI). In the series of short demo videos below, ARMie will team up with a series of experts to show you examples […]
In a recent post, I examined how AI-powered manufacturing solutions enable manufacturers to aggregate data from multiple sources and apply AI for better outcomes. Now, it’s time to talk about how. Below, I have some recommendations on how to begin your AI-powered manufacturing journey. Start small, start anywhere, start now Some manufacturers are put off […]
We are delighted to announce that IBM has joined the Global Manufacturing and Industrialisation Summit (GMIS) as a founding partner. The news first broke at Think 2018 – IBM’s landmark conference, and a unique opportunity for IBMers, partners and clients to innovate together. Given our focus on the power of collaboration, it was an apt […]