With the rapid development of artificial intelligence, AI chip is also becoming a core industry. Recently, the semiconductor manufacturer ROHM and Singapore's global research institute microelectronics Institute reached a cooperation agreement, they will jointly develop artificial intelligence chip.

ROHM, a world-renowned semiconductor manufacturer, has partnered with the Institute of Microelectronics ("IME"), a global research institute affiliated with the Singapore Institute of Science and Technology, and will work on a new generation of plants to jointly detect the detection of abnormalities at the sensor nodes Artificial intelligence (AI) chip.

Conventional device anomaly detection generally processes a large amount of information from multiple sensors to a computer or server within the device. The purpose of this study is to carry out the anomaly detection algorithm which has been processed by the server on the semiconductor chip and process the sensor information in the sensor node, thus greatly reducing the overall power consumption of the system and providing the abnormality detection function which can be used for all devices.

In recent years, in the manufacturing plant, "predictive maintenance" of the abnormality "omen" of the apparatus based on the data obtained from the log and the sensor of the apparatus, in order to take precautions and alternative preparations before the occurrence of the failure, PredictiveMaintenance "concept is becoming increasingly popular. This predictive maintenance is widely used in various sensors and sensor nodes used to sense the state of the device, the wireless module used to send the sensor data to the server, and the software used to process the data. Especially in the field of software, in the past on the basis of statistical processing of artificial intelligence (AI) derived technology is increasing. Linear Technology also want to develop the AI chip.

However, to make full use of the results of such software, the need to continue to send large amounts of data from the sensor, the current problem is that for WSN (WirelessSensorNetwork) * 1) wireless communication technology in continuous transmission of large amounts of data communication speed is not enough Can not cope with the future momentum of the rapid increase in the number of sensors.More and more electronic component distributors will supply it.

In this context, ROHM worked with AME STAR's IME to start working on real-time processing of data from the sensor and sending the results to the server's artificial intelligence chip only when an anomaly was detected. Combining ROHM's unique analytical algorithms (incorporated into a large number of artificial intelligence technologies accumulated through ROHM's plants), integrated with the semiconductor circuit technology (using IME and ROHM's good analog circuit technology) to integrate into the chip. (Figure 2) In this way, can be micro-controller and FPGA and other existing processing system can not be achieved in the advanced processing of the sensor nodes, wireless communication network to build more easily, and can greatly reduce the load on the server. In addition, it combines ROHM's sensors, wireless modules, and battery-free technology, EnOcean * 2, to easily set up wireless sensor nodes in all locations. ROHM and IME are designed to contribute to the replacement of future global plants through this research. You may interested on: VS-T40HF40