5 Ways Machine Learning Is Leading to Smarter Manufacturing

What is ML?

ML is the process of teaching computers how to learn from inputted data so they can make decisions or predictions.

For example, you don’t tell ML how to recognize a cat. Instead, you show many pictures with and without cats. Over time, ML learns what characteristics must be present for an item to be correctly identified as a cat. The computer hasn’t been programmed to know what a cat is; it’s programmed to learn what a cat is.

ML is used to solve problems. In contrast, AI is about getting a computer to perform like an intelligent human. ML algorithms are used in AI, but ML is not AI.

How ML is improving manufacturing

Manufacturers are increasingly using ML because the technology can learn what to expect and is highly accurate when predicting outcomes.

John Rossman, managing partner at technology and innovation consultancy Rossman Partners, says ML’s impact on manufacturing will be significant. “The adoption of ML in manufacturing is setting the stage for a future where efficiency, precision and customer-centric innovation thrive,” Rossman explained. “As manufacturers invest in these capabilities, the true differentiator will be in their ability to ‘think big, but bet small’ and develop a systematic approach to the experimentation required.”

Consider the following five current ML use cases that are transforming manufacturing.

1. ML is improving manufacturing general processes.

Manufacturers use ML-based solutions across the entire production cycle. These systems help detect various issues present in current operating methods; such issues include bottlenecks, unprofitable production lines and quality control challenges.

The marriage of ML and Industrial Internet of Things (IIoT) objects — like sensors — presents exciting possibilities. Companies can gain deeper insights into logistics, inventory, business assets and supply chain management. This analysis can uncover high-value insights into manufacturing, packaging and distribution processes.

An excellent example of this is Intel’s use of ML in microchip manufacturing. Its system captures a series of detailed, high-resolution photographs of the thin silicon slices that comprise a chip during production. ML analyzes these images for defects like scratches or bubbles, which may be difficult for human inspectors to detect.

If an anomaly is found, the system can immediately halt the production process. This way, the manufacturing team can identify and fix the issue before producing additional defective chips.

2. ML is transforming product development.

Product development is one of the most widely adopted ML applications. New product plans, designs and improvements are tied to a vast trove of information that must be considered to yield the best results.

ML solutions can help gather consumer data and analyze it to understand demands, uncover hidden needs and detect new business opportunities. This insight allows companies to improve existing products and develop new ones — opening additional revenue streams while reducing the risks associated with launching unique innovations.

ML is especially good at reducing the risks associated with developing unique new products; this is because its insights feed the planning stages, resulting in more informed decisions.

Val Neicu, CEO and co-founder of SmartSKN, shared how this company leverages ML to revolutionize skincare product development. “By analyzing vast amounts of consumer skin data, ML enables us to create hyper-personalized formulations that align precisely with each customer’s needs, reducing the risks associated with launching new products,” Neicu explained.

At SmartSKN, clients use an AI-powered dermoscope to assess their skin condition. ML then formulates a personalized skincare product tailored to each client’s profile. “Our manufacturing system is powered by ML to produce products on demand, minimizing waste and ensuring efficient inventory management,” Neicu shared. “This dynamic approach allows us to respond in real time to customer demands without overproducing.”

This strategy enhances production efficiency while creating a highly interactive customer experience that combines cutting-edge technology and convenience.

3. ML is advancing quality control.

ML can improve final product quality significantly in two ways:

• ML can find product and packaging anomalies. ML helps companies prevent defective products from reaching the market. In fact, according to Averroes, ML boosts defect detection rates by 60 percent compared to human inspections.
• It can enhance manufacturing process quality. Through IoT devices and ML applications, businesses can analyze the availability and performance of all equipment throughout the manufacturing process. These insights guide predictive maintenance, helping extend equipment lifetimes and avoid costly downtime.

For example, Danish brewer Carlsberg uses ML-powered quality control to detect subtle taste differences between beers that human testers often miss. The company collects detailed data on each beer’s aroma and taste profile during production. When the beer is brewed, the ML analysis model can detect flavor inconsistencies or defects much earlier in the process. This approach reduces the volume of sub-par products leaving the factory and ensures customers receive the products they expect.

General Electric offers another example of ML-driven quality control. The company has deployed ML-based tools across more than 100,000 assets in its business units and customers’ operations; it spans industries like aerospace, power generation and transportation. Its systems detect early warning signs of manufacturing anomalies and provide long-term prognostics to estimate equipment behavior and lifespan.

4. ML is enhancing security.

ML solutions operate across apps, operating systems, networks, cloud computing platforms, on-premise systems and other attack vectors. As a result, mobile app, device and data security are critical concerns for modern manufacturers. Fortunately, ML plays a vital role in the Zero Trust Security (ZTS) framework. This approach heavily regulates and limits user access to valuable digital resources and data, helping protect your business’s sensitive information.

ML enhances security by analyzing user behavior and access patterns. It evaluates how individual users access specific protected information, which applications they use and how they connect to them. By enforcing a strong perimeter around digital assets, ML can determine who accesses what and detect anomalies that quickly trigger warnings or actions.

Manufacturers must take security seriously. According to IBM’s X-Force Threat Intelligence Index, manufacturing has been the most attacked industry for three consecutive years; it accounts for over 25 percent of security incidents. This sustained targeting of the manufacturing sector highlights the absolute need to implement preventative and robust measures.

5. ML is fueling robot technology.

AI technology allows robots to handle routine tasks that are complex or dangerous for humans. With ML capabilities, robots can go beyond the assembly lines they once were confined to and tackle more intricate processes.

That’s precisely what KUKA, a Chinese-owned German manufacturing company, is achieving with its industrial robots. KUKA aims to create robots that can collaborate with humans, enhancing productivity and safety. The company is bringing its robot, LBR iiwa, into the fold. This intelligent and adaptive robot is equipped with high-performance sensors that allow it to perform complicated tasks while working alongside humans; this way, they can learn how to improve their productivity.

Dr. Peter Gorm Larsen, vice head of section at Aarhus University and coordinator of the RoboSAPIENS project, is pursuing many of the same objectives.

“Adaptive robots usually behave by continuously monitoring their environment, analyzing the data collected, changing [their] plans, if needed, and then executing the new plans, accumulating knowledge as it goes along,” Larsen explained. “This is called a MAPE-K (Monitor-Analyze-Plan-Execute-Knowledge) control loop.”

However, Larsen emphasized that this process lacks a step to verify whether certified guarantees of safety and trustworthiness are maintained during behavioral changes. “The future of safety and trustworthiness of automated industrial robots comes when the robot can change its behavior while maintaining — or even increasing — its expected performance and staying at least as safe and robust as before it made the change,” Larsen noted.

Larsen is currently working on the aforementioned RoboSAPIENS project to develop the underlying technology to make adaptive robotics safer and more efficient. These advancements will help reduce human-related errors, boost productivity and add value across the entire manufacturing chain.

The future of ML in manufacturing

ML and AI manufacturing applications continue to attract the attention of investors, entrepreneurs and product developers. Here are some key trends shaping the future:

• TinyML: One recent exciting development is the introduction of ML to small devices: the so-called TinyML. These low-powered devices are easy to attach to production equipment and can make decisions on machine performance or maintenance needs based on the data they receive. This technology promises to reduce delays and speed up the production process, improving overall efficiency.
• AutoML: No- and low-code approaches are making their way into manufacturing ML through AutoML; it simplifies the process of building and deploying ML models without needing extensive technical expertise. Factory workers and engineers closer to the manufacturing process will be able to create and refine tools like maintenance prediction and quality control solutions without the added delay and expense of calling in outside experts.
• Generative AI integration: Dustin Johnson, CTO of advanced industrial analytics consultant Seeq Corporation, highlighted the potential of generative AI in ML. “[Generative AI] can automate complex design processes and the use of ML for real-time supply chain optimization,” Johnson explained. “These innovations enable manufacturers to respond swiftly to market changes and operational challenges.”
• Explainable AI: One concern about ML and AI is the lack of transparency in how they produce results. Explainable AI addresses this by making it easier for users to understand how a system makes decisions — for example, in quality control or supply chain management. With this increased clarity, users gain more trust in a system’s results and are better equipped to interpret and optimize its programming.

ML’s future will rely on skilled personnel who can maximize the technology’s potential. “Data integration and the need for skilled personnel persist,” Johnson cautioned. “Companies need to ensure data quality throughout each project and create a culture that embraces technological change to see the greatest returns.”

Expect further increases in productivity and decreases in equipment failures over the coming years. As ML adoption becomes more widespread, experts predict it will be ubiquitous in manufacturing by 2030.

Nacho De Marco contributed to this article.