October 4, 2018

AI Makes PLC and Robotics More Flexible

					AI makes more flexible

ABB already uses self-optimizing data analysis for painting robots. A networked sprayer equipped with a sensor is connected to the Ability Cloud platform for the painting robots, which enables intelligent, real-time diagnostics and accurate paint control.
Source: ABB

"Preferred fields of application for artificial intelligence exist in industrial environments everywhere where data is generated, which — if properly analyzed — can provide information on machine conditions," Hartmut Pütz stated, Factory Automation President for Europe at Mitsubishi Electric. A typical use case is predictive maintenance, for example. "Using the appropriate sensors, we can record machine conditions in real time and use AI technologies to derive recommendations for action for the machine operator."

Consequently, AI technology is to be integrated into industrial control systems such as CNC and PLC at Mitsubishi, Mr. Pütz commented. In this context, it is a question of being able to adjust control algorithms automatically due to a changed system state. Until now, control algorithms have always been stable, and when the machine state changes, the machine cannot automatically switch to the optimal position. "That is going to change," Mr. Pütz stated. In the future, we want to record sensor data, analyze it using deep-learning algorithms in the PLC and automatically correct the control algorithm based on this analysis.

AI will become an integral part of the PLC

Omron is working on very similar things. "We have an AI machine controller that is the first of its kind to integrate AI processing capabilities and PLC functionality," Dr. Klaus Kluger, General Manager of Central Eastern Europe at Omron Electronics, emphasized. Adaptive algorithms are already an integral part of machine control for the AI controller. This "artificial intelligence on the machine for the machine" learns from real-time data about the movements of a plant in normal operation and can then detect early system failure at the first sign of abnormal patterns.

The AI solution generates the greatest effect when used on the bottlenecks of the respective production process, in which the AI controller increases overall equipment effectiveness (OEE). Testing with pilot customers suggests that the OEE will increase in the single to double digit percentage range, according to Omron. The market launch of the AI controller is scheduled for 2018, initially in the high-end segment, where the necessary CPU computing power is available. "In a few years, AI will be an integral part of the PLC as safety already is today," Tim Foreman, European R&D Manager at Omron, is convinced.

Omron illustrates the potential of machine learning with its table tennis robot Forpheus in a playful way. Dr. Kluger: "Thanks to its adaptive algorithms, the table tennis robot can adapt to its table tennis partners." To this end, Forpheus played table tennis with hundreds of people and learned their behavior patterns. Now, Forpheus can predict the behavior of the players and the trajectory of the ball. In the current version, the table tennis robot is even able to learn from its own mistakes.

Production will change enormously

Robots capable of learning are also in the focus of attention for Kiyonori Inaba, General Manager of the Robot Business Division at Fanuc. He is convinced that only 20 percent of the possible applications are automated. "Eighty percent of automation potential is still unused," Mr. Inaba said. To exploit this potential, collaborative robots (Cobots) have to be combined with AI. Mr. Inaba: "The robots need to get smarter and not just mindlessly handle processes that have been programmed." They have to perceive their environment by collecting environmental and sensor data via the IoT and processing it via AI. In that way, the robot can adapt to the respective situation and decide which task to do and how to do it. Mr. Inaba: "Collaborative robots, IoT, and AI will enormously change the manufacturing world."

Millions of savings in the automotive industry

ABB already uses self-optimizing analysis for painting robots. A networked sprayer equipped with a sensor is connected to the Ability Cloud platform for the painting robots, which enables intelligent, real-time diagnostics and accurate paint control. By monitoring the condition of key spraying components and parameters such as acceleration, pressure, vibration, and temperature, paint application efficiency can be increased by up to 10 percent during the painting process. This should save millions in the automotive industry.

Challenges for the future

As a result, the future of AI in robotics and automation is very promising. However, there are still challenges, as demonstrated during an AI panel discussion at the automatica Forum. For example, Dr. Olaf Munkelt, Managing Director of MVTec and CEO of VDMA Machine Vision, emphasized that you have to put a lot of preliminary work into the data, so that machine learning also delivers reasonable results. Mr. Munkelt refers in this context to the old computer science piece of wisdom: "Garbage In, Garbage Out". In other words, if you feed a calculator or algorithm with non-meaningful inputs, then the result will not be meaningful either.

In addition, the AI panel discussion also showed that the dream of a self-learning, all-encompassing super-AI is still very, very far away. Dr. Wieland Holfelder, Head of the Google Development Center in Munich, therefore primarily sees very narrowly defined use cases for AI in the near future. However, AI provides very concrete benefits there. "We will not have the one AI all-rounder. But we will have a lot of little masters, who make life much easier in partial areas. "