MapR Technologies, Inc., the industry’s leading data platform for AI and analytics, has announced that Edwards Ltd., a world leader in vacuum and abatement solutions, is using the MapR Data Platform to develop semiconductor industry solutions to analyse data from its equipment within customer sites to deliver predictive maintenance and near real-time anomaly detection as part of a major step towards an Industry 4.0-based smart connected portfolio.
Edwards is a leading developer and manufacturer of sophisticated vacuum products, abatement solutions, integrated solutions and related value-added services. Edwards products are critical sub-systems to manufacturing processes for semiconductors, flat panel displays, LEDs and solar cells; are used within an increasingly diverse range of processes. Edwards maintains 30 international sites, employing over 4,200 staff serving all major global semiconductor manufacturing customers.
MapR will provide a core technology in a major shift for Edwards that embraces real-time data analysis from its equipment to help its customers improve quality and yields in vacuum-intense semiconductor manufacturing processes.
As David Hacker, Strategic Marketing Manager for Edwards, explained: “The manufacturing of modern semiconductors include up to 1,000 process steps with a significant and increasing proportion requiring vacuum and abatement systems and a typical plant may have Edwards’ assets measured in the 1,000s as critical sub-systems for these processes.
“Our vision is to have all of these assets delivering near real-time information on their performance and the detection of anomalies that may occur with a deeper level of scrutiny. This allows us to create more evidence-based, relevant information on quality of vacuum service to enhance the knowledge provided to customers so that they can quickly adapt to any situation to deliver more efficiency and better quality of process outcomes. Along with this, the data will enable increased and predictable overall equipment availability along with ‘adaptive triggers’ for scheduled maintenance.”
Edwards chose MapR to enable the capture and processing of diverse data from a wide range of sources that is instrumental in helping it build the machine learning models that will underpin many of these smarter systems. “Our current and medium term aim is to include sensor data to enhance the intrinsic Edwards asset data and to be able to make this service available to other critical sub-system providers to create a more granular and broader understanding of the environment that will ultimately lead to adaptive control based on data analytics,” said Hacker.
“We have chosen MapR following a standard due-diligence method and the experience of the openness of MapR in providing open exposure of their customers to us and thus permitting us to form a preview of what it will be like to work with MapR,” said Hacker. “As our industry sector is still largely focused on-premise solutions for its data, the ability to deliver on premise, hybrid and cloud implementations that all have an intrinsic and universal security model offers us a seamless future migration as and when this industry sector opens up to that transition.”
“Our immediate needs also include the potential to interface seamlessly to other platforms within our customers’ sites so MapR’s provision of comprehensive API’s is vital, but this is also important for us internally as we need to vertically integrate to Edwards’ enterprise resource planning and other related digital services,” Hacker added.
The first deployment of the MapR platform is currently underway in a joint project between Edwards and Fraunhofer-Gesellschaft, in a cooperation that involves research, development and evaluation of specific sensor and IIoT technologies, with the aim of generating and securing high-quality data to feed advanced analytics. Key is the research on the correlation between process and pump behaviour to be piloted in the semiconductor cleanroom environment of Fraunhofer EMFT with a key role played by this new technology stack based on the MapR Data Platform.
Novel Machine Learning techniques will be used to detect anomalies in sensor data. For this purpose, data fusion of several sensor data for combination and pattern recognition, as well as algorithms for detection of specific states of instability are needed. The aim is to improve the process of predictive maintenance by exploring new machine learning algorithms fed with sensor data from above and below the cleanroom floor, to predict future performance and anomalies.
A secure connection will be implemented to allow data, machine learning models and equipment status to be exchanged in real-time between secure locations: on premise at the Fraunhofer EMFT CMOS line and remote in the cloud or Edwards HQ. As well as reducing response time for service, this simplifies analytical model development and deployment. The aim is to demonstrate a new reference architecture, consistent with the principles of RAMI 4.0, but tailored to meet the challenging demands for data and IP security imposed by the semiconductor manufacturing sector.
Jim Stock, VP of EMEA for MapR, added: “Pioneers across many industries recognise that working closely with suppliers, partners and customers is vital to deliver new insights and breakthrough services. Our implementation with Edwards highlights the benefits of a trusted platform to enable multiple partners to seamlessly collaborate on data intensive projects in areas such as the Internet of Things (IoT) and Machine Learning, while retaining the highest levels of security and control.”