According to Gartner’s research, up to 25% of organizations surveyed annually since 2019 said they plan to leverage artificial intelligence (AI) within the next 12 months. However, only up to 5% have deployed AI to production. Companies know that generative AI can dramatically improve operations, but inexperience and potential risks postpone the benefits. Predictive maintenance, efficiency, and autonomy drive AI adoption, and more manufacturers are developing AI-enabled devices in support. Sensors, cameras, edge devices, and automation software have traction, and we expect an explosion of new AI capabilities in the coming years. To get a pulse on what’s happening now, Design World recently asked several industry experts about trends they’re seeing in this space. Here’s what those experts had to say.

What new offerings of yours leverage the power of AI?

Caldwell: For the last several years Yaskawa has been eager to partner with various technology partners that leverage AI for vision-guided robotics and other applications. Most recently, Yaskawa announced Motoman NEXT, an AI-based platform with an integrated Autonomous Control Unit (ACU) that allows for more direct integration of AI within the robot controller. Motoman NEXT is primarily intended to be an enabling platform for the development of automation solutions in new markets and for new applications and end-users. The inclusion of an ACU and the more direct partnerships with a variety of leading automation and technology providers should enable a much greater fusion of sensors, technology, and ease of use.

Latino: Festo has a suite of software solutions under the brand name Festo AX Industrial Intelligence. This software uses AI for machine analytics, with a primary focus on predictive maintenance, predictive quality, and compressed air savings.

Rangarajan: HP shares the belief that advancements in AI and its continued integration across the additive manufacturing (AM) digital and physical workflows are key to unlocking the maximum potential of 3D printing technology. As such, HP has made it a priority to introduce new AI-powered offerings that help customers optimize production and bring more compelling product designs to market.

One example is the new HP 3D Digital Sintering software, announced in spring 2023, designed to address the costly challenge of dimensional and volume shrinkage. AI-accelerated simulation convergence and physics-based modeling make predicting, reporting, and compensating for variability of 3D printed parts during sintering easier, which in turn reduces time, cost, and speed of final part production without the need for a build and test approach. For hardware, HP’s 3D Center product offering helps improve OEE by enabling predictive alert recommendations.

Sachdev: Our newest AI-enabled vision sensor is the In-Sight SnAPP. It comes with pretrained edge-learning algorithms that can help engineers get up and running within minutes — no programming and no vision-system experience needed. The edge-learning algorithms can also learn a specific application with only a handful of training images. We’ve had customers deploy it on their lines in less than four hours.

Mayers: We recently launched a new camera for the IDS NXT AI vision system. The intelligent industrial camera IDS NXT malibu features Ambarella’s CVflow AI vision system on chip and takes full advantage of the SoC’s advanced image processing and on-camera AI capabilities. Consequently, image analysis can be performed at high speed and displayed as live overlays in compressed video streams via the RTSP protocol for end devices. Furthermore, an integrated ISP with automatic features (especially for brightness, noise, and color correction) ensures excellent image quality. IDS is the first manufacturer to make this chip available for industrial applications.

Yudilevich: MaterialsZone leverages the power of AI through its Materials Knowledge Center, a secure database for storing materials data. The data within the knowledge center is accessible, standardized, and ready to use for statistical analysis and in AI models. Our Co-Active Visualizer provides statistical dashboards for simplified lab data analysis, and our Predictive Co-Pilot employs AI and machine learning (ML) models to optimize experiment planning and save time and costs. The predictive co-pilot has an explainable AI component that uses the SHAP values method to rank the importance of the different features in the model and visualize and quantify the effect of each feature on the target parameter. These offerings enhance data accessibility, simplify analysis, and optimize lab experiments.

Caldwell: We have been able to leverage AI to compound the benefit of our Yaskawa Cockpit robot health checker. While the Robot Health Checker function within Cockpit can compare a variety of signals from a robot (encoder temperature, hours used, duty cycle, grease life), traditionally, these comparisons were only made with standard data using assumed conditions. Leveraging AI allows for specific installations to improve the quality and accuracy of their ongoing preventative maintenance programs to ensure zero or minimal downtime of their automation solutions.

Varley: Later this year, we’re releasing an optional upgrade to our robots that will improve high-speed positioning and path control for applications needing enhanced performance. As more devices provide feedback that can be fed into AI algorithms, robot performance and the range of addressable applications can both be increased.

What AI are you seeing in industrial applications?

Rangarajan: Businesses across industries today are looking to increase their use of powerful and sophisticated AI models. Manufacturing is no different, with AI already being used to great effect. From material development and product design right through to process monitoring, optimization, and production, AI is poised to help address the growing demand and availability of 3D-printed parts especially.

Latino: Modern industrial installations leverage technologies that are quite inducive to supporting AI. Access to data is paramount, and today’s networking technology makes it easier than ever to access data. This does not mean that only modern installations can support AI. Legacy installations can also benefit but may need additional infrastructure. Connections to edge computers or cloud systems are more common, and AI may utilize one or both for analytics, modeling, and so on.

Mayers: Interest in AI-based solutions is as great as ever, and the nature of inquiries is changing. In the past, it was mainly about the basic possibilities of AI and cameras. Now, the questions center much more on the specific implementation. Our customers have a much deeper understanding of the topic, and therefore we now provide advice at a completely different level. Clients are interested in the technical capabilities of the system and their value user-friendliness. This may be because AI-based camera solutions are not only used by image processing experts, but by all kinds of users — which is exactly one of the reasons AI vision is so attractive.

Where do you see the most promise for edge AI/ML software?

Halstead: AI is very good at finding patterns in large amounts of data, but it’s not very good at creating a new concept. So, a factory with lots of sensors feeding data that can be analyzed will benefit. Creative tasks? Not so much.

Mayers: There is a high demand for edge AI/ML capabilities, and we are committed to answering this with systems such as IDS NXT and the new IDS NXT malibu camera. These systems exemplify technological advancements with their integrated AI, providing realtime, on-site image processing and decision-making. This is a game-changer for industries requiring immediate data analysis, such as in automated manufacturing, where split-second quality control decisions are crucial. Additionally, the IDS NXT malibu camera’s ability to stream compressed video in realtime is a significant stride in efficient data management and processing.

Varley: For the robotic products that are manufactured by Mitsubishi Electric Automation, we offer the Smart Plus card. This allows access to AI-driven features, such as Predictive Maintenance and Enhanced Force Sense Control. The predictive maintenance function has benefited from years of data analysis to make our wear calculations as accurate as possible. Both AI-driven functions lead to better and safer machines. By using the results of the predictive maintenance function, users can keep the robots running at peak efficiency and reduce the risk of machine crashes that can result from a maintenance-related problem that wasn’t discovered by traditional preventive maintenance schedules. Enhanced Force Sensing control allows a wider range of applications to be automated and potentially dangerous tasks to be taken care of by robots instead of operators.

Dengel: AI/ML can be used to set up factory floors for lights-out operations. This increases productivity and allows for labor to be redeployed into other roles. AI can also be used to simplify and customize user experiences to their unique needs.

Caldwell: Overall, the most promise for AI/ML software is expanding automation beyond the traditional dull/dirty/dangerous tasks where it has traditionally been focused. Historically, automation has been targeted at highly repeatable work, and to great success. But the next wave of automation coupled with AI/ML will be with highly service-based unfulfilling tasks. Food service, textiles, farming, and other unstructured but unfulfilling work will be achievable with smarter automation. There are several benefits, including improved and more reliability, reduced downtime, and long-term margin improvement due to traditional service life measured in decades despite ROI in two to three years.

Latino: AI running on the edge has the possibility to detect faults quickly due to very low latency in collecting sensor data and immediate processing at the field level. Predicting field device failure and product quality as soon as possible is most promising. Some users claim thousands of dollars lost per hour due to a single device failure. Therefore, these AI solutions should have a quick ROI.

Sachdev: AI is having a huge impact on machine vision applications through increased capabilities and better ease of use. We sell vision systems that employ two different types of AI: deep learning and edge learning. Deep learning systems use AI neural networks to learn how to accomplish a sorting or inspection task, distinguishing good parts from bad when defects can be extremely variable, for example. And edge learning systems come with pre-trained algorithms that make training incredibly easy so they can be deployed in minutes, not hours or days. In both cases, the addition of AI means the vision system provides better ROI, so we expect to see AI adoption continue growing.

While AI can improve many types of inspections, it can make a bigger impact in some specific areas, such as inspecting packaging, welds, EV batteries, and food and beverage products. The common thread is that AI excels at learning how to differentiate cosmetic blemishes from functional defects. Croissants coming down a production line, for example, can have a wide level of variability in color and shape that traditional rules-based systems can struggle to tell apart from actual problems. AI systems can learn to recognize that slightly more bent or browned pastries are completely acceptable.

How have you seen AI improve predictive failure and maintenance programs?

Munirathinam: Industrial manufacturing requires precise and efficient production with optimized machining to maximize value-adding activities. AI-based predictive failure maintenance programs enhance the efficiency and effectiveness of maintenance programs while reducing repair costs and contributing to a sustainable future. AI-powered solutions such as Schaeffler’s Medias Xpert mobile app and OPTIME condition monitoring system showcase the potential for growth in leveraging AI to prevent equipment failures, optimize maintenance programs, and promote sustainability.

Latino: There are Festo AX installations today that have predicted failure in industrial applications around pneumatic cylinders, valves, nozzles, and actuators. This has been done by developing models for these systems that include sensor data. Festo, as a designer and manufacturer of industry components, has also developed engineering data relevant to these applications. Festo is in a unique position for AI/ML because of our design, manufacturing, and data science experience.

Munirathinam: AI-powered diagnostics for the full product lifecycle is another application transforming asset management. Many of our industrial customers have approached Schaeffler for assistance in preventing equipment failures, including help with realtime failure analysis to assess the impact as well as recommendations for immediate repair and long-term prevention. This type of diagnosis demands expert industrial experience and end-to-end technical expertise to guide customers effectively. Our AI-based diagnostic tool and mobile app Medias Xpert identifies damage and empowers customers to diagnose bearing failures by just taking pictures on their phone, understand root causes of issues, and receive suggestions for effective countermeasures right in the palm of their hand.

Mayers: AI-based camera systems can process realtime data … and as embedded devices even directly at the source. Utilizing AI leads to faster and more accurate anomaly detection and pattern recognition. Minimizing the delay in transmitting data for processing is also crucial to allow for more rapid identification of potential problems and therefore detecting potential risks and defects before they become an issue.

Munirathinam: For maximizing performance, AI and condition monitoring make predictive maintenance easy. Schaeffler’s OPTIME Condition Monitoring solution consists of wireless sensors, a cellular gateway, and digital services based on proprietary Schaeffler algorithms. Multiple OPTIME sensors are included to monitor pumps, fans, gearboxes, and motors. With this solution, most of these balance-of-plant assets can now be monitored on a continuous basis. OPTIME automatically detects problems, issues the appropriate alarms, and provides information about the possible cause of the problem. Employee safety is also improved as wireless sensors eliminate the need to access hard-to-reach machinery.

Rangarajan: The time required to develop and deploy alarms has significantly reduced with the shift from statistical approaches to AI-based approaches. Coupled with an easy-to-deploy cloud architecture that extracts, transforms, stores, and delivers proactive and predictive alert data from 3D devices in the field, AI is improving the customer experience with proactive repairs and enhanced OEE for end users.

Meet the experts

Chris Caldwell | Product manager – material handling • Yaskawa Motoman
Frank Latino | Global product manager — electric automation • Festo
Arvind Rangarajan | Global head of software and data • HP Personalization and 3D Printing
Gian Sachdev | Marketing head – Americas demand generation • Cognex
David Mayers | Sales director • IDS Imaging Development Systems Inc.
Ori Yudilevich | CTO • MaterialsZone
Patrick Varley | Product marketing manager — robotics • Mitsubishi Electric Automation Inc.
Richard Halstead | President • Empire Magnetics Inc.
Brian Dengel | General manager • KHK USA Inc.
Karthikeyan (Karthikk) Munirathinam | Director of digitalization • Schaeffler Americas

---

Filed Under: AI • machine learning, Trends
Tagged With: Yaskawa, festo, cognex, hp, Schaeffler, empiremagnetics, ids, mitsubishielectric, khkusa, materialszone
 

---