Due to state restrictions to protect Hoosiers from exposure to COVID-19, Conexus Indiana is hosting its regularly scheduled Emerging Technology Showcases virtually. The article below is an excerpt of a webinar hosted on April 29, which can be viewed along with the other showcases in its entirety here. The next showcase on May 27 is titled “Connected Workplace Health Analytics” and will feature LHP Engineering Solutions. Register here.
Historically, manufacturers across all sectors have strived to eliminate waste in production processes and keep them “lean.” In today’s advanced manufacturing landscape, businesses are increasingly turning to Artificial Intelligence (AI) to accomplish this waste elimination. This process begins by collecting data through the Internet of Things (IoT), storing it in a central location like the Cloud, and analyzing that data with AI technology. By learning from the data, AI technologies like those offered by showcase host Mariner will be able to anticipate and respond to events and outcomes, without human intervention.
Case Study: Quality Assurance
AI has had a large impact on the quality assurance operations of businesses able to implement such technology. For example, over the past 30 years, machine vision technology has been the primary method fabric companies use to determine defects in bolts of fabric. This technology functions by collecting images of the fabric, applying a contrast to the photo and using an algorithm to determine whether that contrast is the result of a defect. However, due to the limits of those technologies, they can have a false-positive rate of up to 40%, which was the case for a Mariner client. Such high false-positive rates resulted in a dramatic increase in the number of human inspections needed to ensure perfectly good fabric was not accidentally wasted.
In the past five to 10 years, machine vision technology has been largely upended by AI deep learning technologies. Rather than applying contrast and using an algorithm, the deep learning technology uses training it has received to automatically detect imperfections. This learning functions by taking a large cache of images, classifying the different types of markings that might be present on a piece of fabric and teaching the AI which type is fit for distribution and which is a defect. The AI processes images while on the job, learning what is correct and what is not. In the case of Mariner’s client, this implementation resulted in a 97.7% accuracy rate to determine defects in the fabric. Such a drastic difference in accuracy allowed the client’s workers, who would otherwise spend their time on re-inspecting fabric, spend their time on tasks more valuable to the company.
For more information on AI technologies like deep learning, please enjoy the full webinar here.