Machine Learning Comes to the Factory Floor
The Sysmac AI controller can identify abnormal machine behavior without being explicitly programmed to do so.
A new automation controller from Omron Automation Americas (omron247.com), the Sysmac AI, uses machine learning capabilities to automatically collect and analyze data from automation equipment, then use the data to recognize abnormal behaviors and respond with compensatory programming and predictive maintenance strategies.
Let’s spell that out: The controller can identify abnormal machine behavior without being explicitly programmed to do so. How? It creates data models based on its analysis of continually reported machine process data. In other words, it figures out what the normal behavior is and monitors the machine status based on that knowledge. Then, if an anomaly is detected, it not only reports it but also writes an operational program change that can compensate for it in the near term—possibly saving a workpiece and/or the automation equipment from immediate damage—and calculates the optimal time for maintenance.
This machine-learning capability saves the users from the time and cost of either developing their own analytics and program-optimization capabilities or implementing a costly third-party cloud-based system that would be slower than this solution. The time needed for the Sysmac AI to detect an anomaly and generate a new instruction that will compensate for it is, according to Omron Automation Center Director Mike Chen, fewer than 10 milliseconds.
And one of the reasons the system can work so fast, according to Chen, is because it is not a cloud-based Industrial Internet of Things/Industry 4.0 solution. That is, the data is not being routed to a distant data center or up to the cloud to be accessed and analyzed with the rest of a factory’s data. Simply, recursive, two-way communication is happening between the automation equipment and the controller. “In this way, the data integrity can be relied upon for ultra-high-speed analysis within our integrated machine learning engine for anomaly detection,” Chen says.
Topology optimization cuts part development time and costs, material consumption, and product weight. And it works with additive, subtractive, and all other types of manufacturing processes, too.
Nowadays in the U.S. market, vehicle manufacturers pretty much are all committed to producing crossover utility vehicles rather than their predecessor type, the sport utility vehicle.
Automotive manufacturers are meeting CAFE fuel-efficiency standards through lightweighting, which requires simulation software for design engineers.