Digital technologies are transforming industrial production processes, offering manufacturers greater flexibility, increased productivity and revenue, and better quality products. This fourth industrial revolution—or Industry 4.0—will see ‘smart’ manufacturing become the norm, with technologies such as IoT, AI and ML capturing and analyzing data in real time, and making decisions autonomously in real time. Gartner predicts that, by the end of 2023, one in five industrial equipment manufacturers will support remote industrial IoT capabilities, and IDC FutureScape says that by 2026 70% of all Global 2000 companies will use AI for operational decision-making. And according to the latest IoT Signals report from Microsoft, the 91% of manufacturing organizations who are IoT adopters say they benefit from increases in operational efficiency and production capacity, and reductions in human error.
In this blog post, we round up how some of the main advances in ML and AI are modernizing manufacturing.
Making factories smarter and more efficient
Predictive maintenance is one of the key components of this smart industrial revolution, ushering in a new world where companies are no longer limited to managing equipment based on age or usage. Powered by IoT, ML, and AI, it gives manufacturers the ability to predict major equipment failures or breakdowns that will significantly impact the business before they happen, reducing downtime and the costs associated with unanticipated maintenance. Telemetry from IoT devices and sensors provides real-time visibility of equipment status, and anticipates problems early on, thus preventing costly failures. Instead of putting pressure on engineers to repair or replace equipment, manufacturers can now predict and prevent issues before they cause problems.
Cloud-based IoT service solutions such as Hitachi Solutions’ IoT Service Hub, which combines Microsoft Azure and Intel processors, contain autonomous predictive service capabilities that intelligently predict failures, initiate corrective actions, and facilitate the repair process—all to prevent the failures before they occur. And, by gathering data-at-the-edge then applying AI to provide real-time visibility and alerts, it helps organizations predict environmental and machinery issues and deal with them before they potentially cause a safety hazard, protecting workers’ well-being and ensuring safe equipment and working environments. In addition, fleet and route optimization and workplace safety analytics use IoT devices to monitor tank levels, driver behaviour, and fuel consumption, using the data to improve equipment turnaround times and increase efficiencies and reduce costs.
Using computer vision to improve defect detection
False rejection is a costly issue for manufacturers, but it is an area where IoT and AI can help. It’s when a ‘good’ product is erroneously discarded from a production line during safety and quality inspection. In the worst-case scenario, falsely detecting contamination could result in the rejection of a whole batch rather than the individual product alone, or significant wastage, and costly product recalls and disposal. By improving false detection rates, IoT and AI helps manufacturers eliminate these persistent ‘pseudo-defects’ and the resulting, unnecessary scrapping, and can significantly reduce their Total Cost of Quality (TQC) and reinspection labor costs.
The Spyglass Visual Inspection solution from Mariner, for example, combines Azure IoT Edge and Intel Xeon processors to create a deep learning algorithm that outperforms the capabilities of human inspectors. Take the example of a global manufacturer of automotive glass. Traditional machine vision systems might incorrectly label things like water droplets or dust as defects. False positives such as this meant that they were losing over $30 per unit over 40 production lines. Spyglass Visual Inspection enabled them to augment their existing vision systems in order to detect defects more accurately at high speed and in large volumes, helping achieve over $1 million in quarterly savings.
Optimizing production yields, quality, and waste reduction
Extruders, robotic, chemical, and other production processes are used for countless manufacturing processes, but traditional approaches can create problems which cost time and money to resolve. Extruders age, raw material characteristics vary between batches and suppliers, and other environmental variables change, which present a significant challenge for traditional control systems. Autonomous AI systems control or propose new parameters in real time, based on expert training to ensure optimum production yield, quality, and waste reduction.
Similarly, it takes time to programme robotic control systems and—once deployed—these programs can only support limited environmental changes, which means that a robot can be thrown off course easily and require human intervention. Another challenge in these industries is that the labelled data required to build a machine learning model simply isn’t available, and it’s impractical to run hundreds or thousands of tests on a live device or process.
Production Yield Optimization from Neal Analytics solves all these problems. Using Machine Teaching, process simulation, Deep Reinforcement Learning, and IoT and Edge computing, AI agents are trained to support operators and can control parameters in real-time to improve yield, quality, and waste. These autonomous systems use AI to make the robotic process more effective and resilient to environmental changes, allowing production to be accelerated and normalized. And by using reinforcement learning on simulators they bypass the labelled data limitation, instead leveraging the expertise of process specialists themselves.
It’s exactly what Neal Analytics did to ensure that leading food and beverage manufacturer PepsiCo created the perfect Cheetos corn puff every time. Working with the Microsoft AI engineering team and the company’s manufacturing experts, they built, trained, and deployed an autonomous system, leveraging the Microsoft Project Bonsai platform to create an AI agent that helps operators optimize the yield of the Cheetos extrusion process.
Further uses of IoT in manufacturing
Predictive maintenance, defect detection, and yield optimization are just three of the many uses of IoT, ML and AI in the manufacturing industry. Other applications such as advanced demand forecasting helps predict demand at individual sites—and even at SKU level—by leveraging data analytics and machine learning techniques. By analyzing external and internal factors that influence demand, including how they influence each other, organizations can optimize their supply chain and business operations as well as develop strategies to maximize sales.
One other, relatively new, application is assembly verification, where AI is used to read and understand bills of materials, parts specs, and CAD and other installation drawings, and then correlate those documents and drawings into output that helps to ensure that article assembly happens completely and accurately. Spyglass Assembly Verification from Mariner uses Azure and Intel processors to keep its eyes on the article as it moves through the production process, ensuring that it sees what it expects at each station and alerting workers when it doesn’t. The result is a lower Total Cost of Quality (TQC) by avoiding costly re-positioning or re-working of high-value items, and can produce even greater ROI when used to stop improperly assembled articles from reaching value-add processes like paint application.
Building a connected future for your business
These are just a few examples of how the combination of IoT, ML and AI are modernizing manufacturing. They show how, together, Intel and Microsoft are enabling intelligent solutions that are changing the industry landscape, enabling their ecosystem of partners—including Mariner, Neal Analytics, and Hitachi Solutions—to create turnkey solutions from edge to cloud, securely and reliably.
Learn more about how Microsoft and Intel technologies work together to improve customer experiences, streamline operations, improve product quality and employee safety through pre-built, proven industry-specific solutions from trusted providers at The Intelligent Edge.
Posted at https://sl.advdat.com/37wopKjhttps://sl.advdat.com/37wopKj
