AI optimises factory automation processes in real time
Mitsubishi Electric and Japan’s National Institute of Advanced Industrial Science and Technology (AIST) say they have developed an AI (artificial intelligence) technology that makes real-time adjustments to factory automation (FA) equipment while it is running. As well as eliminating the need for time-consuming manual adjustments, the AI estimates the confidence level of inferences regarding factors such as machining errors, and then controls the FA equipment based on suitable levels of confidence.
The partners predict that the technology will lead to more stable, reliable and productive operations, especially in agile manufacturing.
To optimise conditions in such manufacturing, parameters such as operating speeds need to be adjusted frequently. However, doing this by hand is laborious and time-consuming, resulting in decreased productivity.
In response, Mitsubishi and AIST – who have been collaborating on AI since 2017 – have developed a technology that uses AI to predict variations in manufacturing processes, such as changes in workpiece shapes as they are being machined, and then adjusts the FA equipment’s operation automatically in real time. In addition, the confidence levels of the AI inferences are indexed and the FA equipment is controlled to ensure high reliability and productivity.
The two organisations cite several examples of how their AI technology is being applied in practice:
• Mitsubishi has used it to develop a method for estimating loads on robotic arms. Various load parameters are used to calculate acceleration and deceleration speeds. The AI function quickly infers load values using information about the robot, such as its motor currents, joint angles, and so on. Simultaneously, the confidence levels of the inferences are calculated. The robot’s acceleration and deceleration are adjusted based on estimated values and confidence levels. A validation test that compared differences in robot motion when using and not using the load inferences, found that robot operating times were reduced by 20% when inferences were used.
• In a second application, an engraving EDM (electric discharge machine) is being adjusted automatically. The EDM positions an electrode close to the workpiece and generates an electric discharge to perform engraving. Debris produced during the machining process must be ejected. As processing proceeds, the amount of debris increases, so needs to be ejected more often. Mitsubishi is using AI to learn the state of debris ejection and to adjust the ejection frequency automatically. Machining times have been reduced up to 23% compared to processing without AI.
• A third example is an AI-based error-correction system for CNC cutting machines. The AI estimates changing machining errors – the difference between the cutting machine’s current position and a command value – to enable correction even during dynamic machining. Tests have shown that machining accuracy can be improved by 51% compared to using error correction that is not supported by AI.
According to Mitsubishi and AIST, the AI technology has several key attractions:
The Mitsubishi/AIST AI technology will provide real-time optimisation of automation equipment
• It is fast, achieving high-speed inferences for dynamic control of FA equipment. In conventional manufacturing, skilled workers must adjust operating parameters to achieve required specifications, such as accuracy levels. The AI technology simultaneously performs high-speed inferences and equipment control for real-time FA operation. It can achieve high-level inference accuracy while simultaneously guiding FA equipment control.
• It can adapt to constantly changing production factors. Workpiece shapes change during manufacturing and this can lengthen manufacturing times or reduce processing quality. Changes can vary for each workpiece, making it difficult for FA equipment to learn in advance. The new technology allows the AI to learn work factors while operating the FA equipment and then to make real-time adjustments. It also formulates physical phenomena, such as friction, and incorporates these mathematical expressions to adapt to constantly changing processing factors.
• It performs adjustments reliably. AI inferences must be reliable to ensure that real-time control of FA equipment leads to stable product quality and efficient processing. Mitsubishi’s new algorithm calculates the confidence level of inferences by learning the characteristics of each process and each target device, ensuring high reliability.
Mitsubishi expects to incorporate its Maisart AI technology increasingly into its FA equipment and systems to improve manufacturing productivity “significantly”.