23 Jul 2024

AUTOMATION FOR MANUFACTURING

Module adds AI to industrial controls and I/O

The bland exterior of Siemens’ artificial intelligence module gives no indication of the power that it could add to industrial applications.

Siemens has developed a module that adds artificial intelligence (AI) capabilities to its Simatic S7-1500 controllers and the ET 200MP I/O system. The S7-1500 TM NPU (neural processing unit) is equipped with an Intel Movidius Myriad X Vision neural network chip, which is used in combination with a trained neural system on an SD card.

The module can process data from a connected sensor system – such as a camera or microphone – or from the CPU. Using machine-learning algorithms, for example, it can guide robot systems or perform visual quality checks in production plants. It is said to result in efficient, “human-like” behaviour.

In pick-and-place applications, for instance, it can help mobile robots to recognise, pick out and place components that are lying randomly in a crate. It quality control applications, it can transfer expert human knowledge about parameters such as the consistency, colour and quality by continuous training of a neural network with image data from a connected camera, for example.

Data from sensor systems (connected to the module via USB 3.1 interfaces or a Gigabit Ethernet port), as well as information from CPU programs, is processed using neural networks. The results are then evaluated in the CPU program.

Intel’s Myriad X VPU chip is claimed to be the first in its class to have a dedicated hardware accelerator for deep neural network structures. An integrated image processor combined with the neural network processing is said to make the Myriad X a “trailblazer” for computer vision applications. Siemens predicts that the chip will open up new industrial automation applications by speeding up image-processing and evaluating data rapidly using trained models.