Sage Automotive increases efficiency and production using Mariner’s Spyglass Visual Inspection and Microsoft Azure to monitor defects
"There's just like a million benefits, honestly, to having a map of every roll of fabric we have that shows us where every defect is and the false images. It is so advantageous for us because we're able to run it through a frame faster and trust that it will only stop on the true defects. It is increasing efficiency, production rates and quality and standards within the plant. With all this data, we are able to improve across the board.”
Sky Williams, Process Improvement Specialists / Sage Automotive Interiors
Product quality situation threatens margins
Sage Automotive Interiors, headquartered in Greenville, South Carolina, is a global manufacturer of high-performance automotive interior fabrics and has been operating for more than 70 years. They needed to solve a quality problem with its material. Sage worked with Microsoft partner Mariner to deploy Spyglass Visual Inspection (SVI), an AI, IoT, and deep-learning solution that utilizes Microsoft Azure IoT Hub and other Azure services. Sage started with a proof of concept, and the digital transformation leader at the company was pleasantly surprised at how easy it was to build the first deep-learning model, then scale with Azure. Using Mariner’s Spyglass Visual Inspection product, Sage is now able to run its production lines faster, increasing efficiency and quality standards across the board.
In 2019, a sales manager at Sage discovered a problem with material quality.
Ashton Paoletti, Director of Information Technology - Americas at Sage, said, “Some interior fabrics are lower-cost, higher-volume, lower-margin product, so any quality issues would eat into what little margin there was. The sales manager was previously a plant manager and knew we had a camera system in place and asked me to figure out if there was a way we could do something more with the images it produces to help improve quality.”
Paoletti reached out to his Microsoft representative, who facilitated an introduction with Mariner.
Based in Charlotte, North Carolina, Mariner delivers Deep Learning-powered solutions that help manufacturers reduce their costs of quality both internally and externally. Spyglass Visual Inspection, its flagship product, provides enhanced defect detection in situations that traditional machine vision systems have trouble accurately handling, and also offers Spyglass Assembly Verification, a solution that ingests engineering drawings and BoMs from a customer's ERP and then uses Deep Learning and visual inspection to ensure that highly complicated or customized articles are properly assembled. In a testament to its successful work, Mariner won the Microsoft 2020 IoT Partner of the Year award.
When Mariner spoke with Paoletti, Mariner suggested their Deep Learning approach. But this was Paoletti’s and Sage’s first Deep Learning project, and with the travails of their existing camera system, Paoletti said, “If I'd gone to them (the CEO and vice president of manufacturing) and said, ‘Hey, we should try this, because it might work,’ they would've said, ‘There's no way that's going to work.’”
Consequently, Paoletti asked Microsoft to fund a proof of concept in which Mariner built an initial deep-learning vision model from sample images. Paoletti said, “I was actually pleasantly surprised with how well building that first model went, how successful it was, and how easy it was to deploy and scale with Azure.”
After the POC results were presented in a regular manufacturing review, Sage’s VP of manufacturing was convinced, and he persuaded the CEO to approve a three-year, multi-site Spyglass Visual Inspection subscription.
Defect detection powered by Azure services
Spyglass Visual Inspection, available in the Azure Marketplace, provides real-time defect detection. It uses Azure NC-series virtual machines for data science, along with Azure Blob storage, Azure SQL Database, Azure IoT Edge, Azure IoT Hub, and Azure Container Registry for other functions. Mariner selected these technologies for their alignment to Spyglass Visual Inspection’s product roadmap drivers of confidentiality, scalability, and simplicity.
Spyglass Visual Inspection is comprised of edge and cloud components, and it utilizes an Azure IoT Edge server installed on-premises. Azure Blob storage is used for storing images and image metadata, while Azure SQL Database is used to report performance. Azure IoT Hub connects the Azure IoT Edge server to the cloud, and Azure Container Registry is used to manage the container lifecycle from cloud to edge.
Working with the vision model
Sky Williams, Process Improvement Specialist at Sage, was given the responsibility of integrating Spyglass Visual Inspection into Sage’s production and saw it as an opportunity to help her team of graders.
“If they are able to use a system that helps them get more yards of material, but also do their job better, they will have more confidence in their job,” Williams said. “My goal is for them to have an easier job and be more productive, thus increasing their pay and having a happier and more sustainable life.”
Williams spent many hours training the vision model by correcting or confirming its defect classification results to improve its performance. Now, instead of spending 100 percent of her time on inspection, Williams regularly adds new training data and spends most of her time “verifying the AI system is functioning the way it was last week and the same as it was functioning yesterday.”
Originally, the graders looked at the fabric as it went by. After initially being intimidated by using a computer screen with images of defects, they are now comfortable with it – a timely transition, as the upgraded line runs too fast for the human eye to detect defects.
Mapping rolls of fabric and tracking defects
Paoletti’s advice for those who would follow a similar path: “It’s not magic. There has to be some sweat equity put into making this work, and I certainly didn't have an appreciation for how much effort it would take for this to be successful. It's not just something you can come in and just casually think, ‘I will go look at these rolls today.’ You can tell Sky is really interested in making this successful.”
Williams said, “It is a labor pain for something new that, in the end, has good results. There's just like a million benefits, honestly, to having a map of every roll of fabric we have that shows us where every defect is and the false images. It is so advantageous for us because we're able to run it through a frame faster and trust that it will only stop on the true defects. It is increasing efficiency, production rates and quality and standards within the plant. With all this data, we are able to improve across the board.”
Williams and Mariner are collaborating on Spyglass Visual Inspection improvements to further increase the graders’ productivity. The VP of manufacturing has also approved an upgrade for root-cause analysis combining Spyglass Visual Inspection’s image analysis with our virtual production manager to notify process leads responsible for adverse trends in real time.
View the original customer story on Microsoft's site here.