AI and collaborative robotics are expected to see impressive growth over the next few years. Manufacturers use robots to meet changing customer needs and fluctuating market demands. But as the digital transformation continues to impact Industry 4.0, we are beginning to see the convergence of automation and IoT in the form of AIoT: the artificial intelligence of things.
We caught up with Jens Beck, Partner Heading Data Management & Innovation at Syntax, to learn more about how AIoT can improve manufacturing systems.
design news: What kind of AI affects manufacturing? machine learning? Education? quality control?
Jens Beck: In general, four types of AI are currently being discussed. The two most important manufacturing implications are reactive machines and limited memory. In a reactive machine, the AI responds to inputs and produces outputs, e.g. B. if the temperature is above a threshold, trigger an alarm. In practice, this would be called condition monitoring. This finds wide application in quality inspection or simply within MES systems. However, this information is not saved. When memory is limited, input and output are correlated and stored to enable predictive use cases or visual inspections. When we talk about anomaly detection to identify outliers or when we use optical systems for visual inspection, we use AI with limited memory.
In the simplest cases, these are simple AI models that have been trained based on the knowledge of the operators or quality controllers and need to be retrained over time. With this approach, you provide the “machine” with good and bad images to teach it what is a good result and what is a bad result. This is particularly important if you want continuous quality control during peak production times, eg cathodes for batteries. In a more complex scenario, you would implement competing neural networks that not only store data, but train themselves as they run.
So where to use AI in manufacturing? Well, predictive maintenance (i.e. predicting when a machine will need maintenance). Predictive Quality is another example, allowing machines to predict an outcome and adjust accordingly based on sensor and environmental data.
Visual inspection is a great use case because it can increase product quality, reduce the manual effort involved in quality inspection, reduce manufacturing time and thus increase throughput. However, this is not the end of the wide range of possible uses of AI in manufacturing. Augmented reality and natural language processing with chatbots can signal operators when they need to increase workplace safety.
DN: How is AI used in collaborative robots? Do the robots communicate with each other? Submit each other’s work?
Jens Beck: Well, collaborative robots are robots that interact with humans, and safety is of course a major concern here. That is why robots on the shop floor are usually kept behind solid fences and interacting with them always means a production standstill.
Now imagine placing sensors around the robot that let it see what’s happening around it. In this scenario, instead of stopping the robot, it could simply slow down its arm or change its movement to avoid harming its human counterpart. Robots could also adapt to the work speed or behavioral pattern of their peers to achieve optimal operation.
All of this requires AI in the background. So it can be said that collaborative robots don’t exist without AI. Again, this would not be the type of AI to become independent, it still remains controlled by humans and is regularly retrained to ensure maximum safety for employees. Of course, an employee can also be another robot, and in that sense the same applies, but with the goal of maintenance reduction and OEE optimization.
Last but not least, AI can also be used to train robots. Let’s assume the robot can replicate typical human movements. Then you could simply record a human doing this movement and project it onto the robot. This technology is there and also uses AI in the background. Of course, this is more likely to apply to areas where such behavior is desirable, such as B. in healthcare, where humanoid robots are increasingly used.
DN: Does AI in manufacturing require the manufacturing equipment to be “smart” equipment?
Jens Beck: If you intend to optimize the maintenance of a machine, its bottom line or OEE, you first need to collect data from that machine. Then correlate them with relevant data from other sources like MES, ERP and historians to get relevant insights and actions.
So the simple answer would be yes. Nevertheless, there are three different generations of machines in workshops around the world. The youngsters – are talkative, polyglot and use the latest slang (or protocols in the sense of IoT). This generation comes smart off the shelf.
The middle generation – is talkative, not fully polyglot and may use some old slang. There are translator solutions for the middle generation that also make them “fully” smart.
The last generation is the “Grandpa” generation – quiet, doesn’t talk much or at all and doesn’t use slang. You can make them smart by retrofitting, i.e. attaching sensors that make them talk. In my experience, this works quite well and, with a few exceptions, provides the insights you need about “grandfather”.
To answer the question, the machine has to be “smart” – yes, but that doesn’t mean you have to make big investments to achieve that goal.
DN: When using AI with factory gear, does it matter which vendors created the gear?
Jens Beck: No it doesn’t and it shouldn’t. Of course, if you bought a brand new machine that comes with an IoT portal as part of the price, you would love to use all your manufacturing machines on that portal. But when all your machines are of the same make and generation, you inevitably encounter obstacles in the workshop.
However, if you look at agnostic IoT platforms, i.e. IoT platforms that are not created by a machine manufacturer, you will find that they are very open about their input options. Their main differentiation lies on the performance and cost side.
DN: Explain the difference between IoT and AIoT.
Jens Beck: IoT is when things communicate with each other, e.g. my alarm clock with my coffee machine; the triggering of the alarm signals the preparation of the coffee machine. AIoT is the world where artificial intelligence helps to make more things talk to each other, i.e. where the output of one thing needs to be interpreted to gain insight that then serves as input to the next thing.
So, in the coffee example, if my alarm goes off, then a camera in the bathroom mirror takes a picture of my face, it notes that I look very tired, so my coffee is made with a double-shot espresso instead of my usual lungo.