A significant number of data and artificial intelligence (AI) projects never make it out of the lab—but operationalizing AI is where businesses can extract some of the most important benefits from the technology. In manufacturing, the use of AI at-scale can streamline work, improve the accessibility of data throughout the company, and enable new maintenance strategies. Manufacturers that can move AI beyond pilot projects and small-scale experiments will see major benefits.
One of the most common applications of Internet of Things (IoT) in industry is operational monitoring. AI can significantly extend the benefits this process can offer. Predictive maintenance with AI improves upon a more conventional preventive maintenance approach by predicting when factory equipment will need repairs.
With this approach, IoT or other networked sensors installed on factory machinery gather a wide variety of data. Typically, this information includes machine vibration, noise, and temperature, but it may also comprise engine lubrication or the differential pressure of a machine’s hydraulic fluid.
This information offers two major benefits: First, it provides real-time data on the health of machines throughout a factory or other site. If equipment is behaving unusually—running hot, moving erratically, or making noise suggestive of a gas leak—site technicians will have instant notice that something may be wrong.
This data also enables the predictive maintenance approach. An AI algorithm, trained on maintenance and operational information gathered by networked sensors, can predict when machines will fail or need upkeep. The predictive approach can improve the effectiveness of a preventive maintenance scheme and minimize the cost by reducing downtime and the number of checks necessary.
According to U.S. Department of Energy data, predictive maintenance can provide savings of 8% to 12% over preventive care. It also reduces downtime by 35% to 45%. Companies relying on reactive maintenance will see savings of up to 40%, and since effective equipment maintenance is one of the best ways to optimize plant performance, businesses may also see improvements to factory output. As the use of predictive maintenance scales across the company, this data will become extremely valuable. AI could uncover relationships between information like environmental conditions and machine performance. With more available data, the algorithm could find more useful and unique insights.
A conventional quality control automation solution requires company developers to manually create rulesets for automated quality control systems. They determine when products are irregular and need to be flagged for review or pulled off the production line. AI can automate the process of creating those rules—both accelerating the development of the quality control solution and potentially providing a more effective ruleset.
Typically, this AI approach uses deep learning. The developer of the AI algorithm begins with a database of products—like a mixture of good and bad ventilator valves. It is trained on this information in a supervised fashion, and, over time, it can determine what makes a valve defective, or at least what surface irregularities suggest a problem.
With enough time and data, the AI can confidently and precisely classify products into groups like “good” or “defective.” Manufacturers that don’t have a large database of product images can use alternative learning techniques like lifelong deep neural networks (L-DNNs). They can create new rules on the fly based on information from the production line. This learn-as-you-go approach enables the use of quality control AI even if existing product data is thin.
The smart quality control approach offers a few major benefits to companies that implement them at-scale. It often provides improved accuracy at a speed not possible with human workers, and the algorithm can also be much more cost-effective once it’s in place. The quality control algorithm can also help manufacturers facing labor shortages reduce the workload that needs to be managed by employees. Implementing AI quality control can also allow workers to move on to tasks that are not as tedious or repetitive and are easy to automate. As with predictive maintenance, large-scale deployment of quality control AI means more data is available, which makes the algorithm better at detecting irregularities over time.
Effective inventory management requires a business to balance two needs—stocking enough items to fulfill demand, while avoiding overstocks that can reduce space and available capital. Predicting demand is essential, but difficult even with good data. Fluctuations can be hard to see coming, and even minor mistakes in inventory purchasing can have significant long-term consequences.
With AI, businesses can develop more effective predictive models—enabling companies to manage the movement of inventory through an increasingly complex global supply chain. For example, AI can be used to streamline the planning of product replenishment and offer more accurate arrival time estimates for shipped goods.
In the same way that predictive analytics can help a business know when a machine will fail, AI can be used to create models that estimate future demand and delivery times based on existing and historical inventory data.
The advantages of operationalized AI can also extend beyond inventory management to related issues, like factory layout. Many manufacturing companies are experimenting with smart asset and inventory tracking. They outfit items and site assets with GPS trackers or RFID labels, which they can use to track things as they move throughout a factory or warehouse.
This information provides managers with a real-time, top-down view of site traffic patterns. If a warehouse or factory has an inefficient layout that is causing bottlenecks or similar issues, they will become immediately apparent. This allows owners to make process changes that can clear up site traffic. As with inventory, AI can be used to identify inefficiencies in factory layout and suggest changes that may reduce traffic or improve throughput. Once the changes are in place, business owners can quickly see how effective they were—enabling rapid experimentation with methods that minimize disruption while maximizing benefits.
AI applications like predictive maintenance, intelligent quality control, and smart inventory management can offer significant benefits to manufacturers. With the right data and AI algorithm, business owners can streamline processes and develop highly effective predictive models—enabling lower maintenance costs and improved demand planning.