Predictive maintenance is just the first step: In the future, artificial intelligence (AI) will revolutionize individual processes and entire plants. But it is still a long way to complete autonomy.
“AI will play an increasing role in production,” Prof. Berend Denkena, President of the WGP – Scientific Society for Production Technology, is convinced. Although AI is not really a new technology, “But today the systems have recourse to better foundations for training machines thanks to the larger amounts of data available and the possibilities for processing and storing these data. In addition, this data availability will continue to increase in Industry 4.0.”
The fields of application for machine learning in production are quite diverse. They range from predictive maintenance to new data-based services to pay-per-use models and all the way to production optimization.
An example of intelligent maintenance is Tünkers. The machine manufacturer from Ratingen fills its machine and systems with a variety of sensors. These data are collected and analyzed in the cloud, initially for condition monitoring. “In the next step, however, the data will also be used for predictive maintenance,” Ralf Görtz emphasized, mechatronics pioneer at Tünkers Maschinenbau. “However, the entire collection, storage and transfer of data would only be a data grave without AI for evaluation,” Mr. Görtz is convinced. “The flood of data cannot be evaluated manually. On the other hand, deep learning algorithms can be used to derive not only the current but also the future state of the machine from the collected data.”
AI now also plays a key role at Trumpf. “Artificial intelligence affects the activities of several hundred employees at our company in various forms,” Dr. Thomas Schneider reported, Managing Director Development of the Machine Tools Division at Trumpf. For example, AI is used in quality control or in AI solutions that make repair suggestions to service staff.
In addition, Trumpf uses AI directly in its machines. The best example is the Trulaser Center 7030 full laser machine. This machine cuts parts from a sheet metal plate and detaches them automatically. Since the sheet metal parts can vary greatly, the machine must remove them from the sheet metal plate in very different ways; otherwise they jam. The machine finds the right strategy for this using AI: If the removal does not succeed immediately, the machine automatically initiates repetitions. Mr. Schneider: “We evaluate the data on successful withdrawals centrally and then transfer the results to all other machines. With customer feedback in the hundreds of thousands, the systems can be perfected continuously.”
The paint shop manufacturer Dürr uses cloud and AI technology for its Ecoscreen Equipment Analytics product, which evaluates robot and process data to make processes in paint shops transparent. For this purpose, Dürr also develops modules that work based on artificial neural networks. “The software automatically learns the optimum process state and registers any deviations. This means that the software will solve problems itself in the future thanks to machine learning,” Dr. Lars Friedrich, CEO of Dürr Systems AG, stated.
Artificial intelligence scores not only in complex contexts such as painting systems, but can also help revolutionize individual processes such as handling tasks. “Industrial handling will be reinvented in the coming years,” Prof. Dr. Markus Glück said, Managing Director of Research & Development at Schunk. Where every single step was previously programmed in an elaborate manner, tomorrow's handling solutions would act much more independently. “Intelligent handling systems composed of flexible grippers and cameras can already be trained intuitively using artificial intelligence methods in laboratory applications in such a way that gripping tasks can be performed autonomously. We will also see rapid progress here in the coming years.”
The Kuka subsidiary Swisslog also has concrete implementation experiences with AI-based gripping with its automated picking solution Item PiQ, which combines a small robot with an intelligent vision system and machine learning functions. The vision system, consisting of a 3D camera and intelligent software, identifies the optimal gripping points for the respective articles and sends the corresponding information to the multifunction gripper of the robot. Thanks to machine learning, Swisslog's SynQ software continuously improves order picking accuracy and gripper performance.
Consequently, will we soon also have self-optimizing systems in production halls?
Professor Denkena believes we should be cautious with respect to such euphoria: “As in the case of motor vehicles, we assume that assistance systems will first be used to support production systems.” These could then gradually enable a certain degree of autonomy. “Our WGP institutes have demonstrated in various research projects that a certain degree of autonomy in machining processes is possible with integrated sensors and corresponding algorithms. This makes it possible to shorten the running-in and setting-up processes significantly. I expect that we will see these systems on the market within the next ten years.”