Semiconductor makers are among the best in the business when it comes to employing automation, data analytics, and artificial intelligence (AI) to their factory processes. Intel is a leader in smart manufacturing, even among its peers.
In a detailed use case, it discusses its use of advanced predictive maintenance techniques to monitor the health of its fan filter units (FFUs) in its factories through deployment of sensors and edge computing. It has reduced manual downtime from FFU failures by 300 per cent over manual inspection, it says.
Elsewhere, in a blog post, it discusses how manufacturers can “grow” predictive maintenance by collecting and comparing data from tools across entire factories, using the FFU study as a template. “Predicting tool failure cost-effectively is achievable, but any IIoT solution should begin with measurable success criteria,” says Robert Colby, principle engineer for IT infrastructure at Intel.
“By understanding the value of what you want to collect and starting small with a single, definable process, an IIoT solution can demonstrate ROI and lay the foundation for measuring the health of other tools and components throughout the factory.”
In this article, Intel discusses its cycle of continuous improvement with Enterprise IoT Insights. Answers are from Colby, along with Steven Meyer, senior principle engineer in Intel’s manufacturing IT division, and Dr Stephen Chadwick, senior principle engineer in its IT team.
How is digital / IoT tech changing the manufacturing industry (on the production line)?
“We have been collecting sensorial data from our equipment for many years, and this data is used to control the tools on-line and is also used for reactive diagnostics. In a few cases we also augment the equipment’s native sensor data with additional sensors”
How are you using digital tech to improve production / operations?
“The sensors we use help control the tools and aid in diagnostics. Our manufacturing is heavy on compute, and we have grown our manufacturing IT over many years to a state where most of the operation runs automatically and the Remote Operation Center primarily handles exceptions.
“We have a combination of central manufacturing execution systems (MES, and execution controls) integrated with equipment and [statistical and advanced] process quality control (SPC, APC, and excursion prevention), which actively re-rout, stop, or adjust equipment parameters and material flow, [plus] edge PCs that control all tools and offline databases.
“This architecture enables the Intel factory to run in high velocity while keeping all process parameters in tight control limits, while responding fast to any changes. The huge amount of data that has been collected enables us continuously learn and compare diagnostics across factories.
“We are continually striving for lights-out automation. This means we look for efficiency improvements in our automation system to improve cost, speed, and yield.”
What is the single most transformative technology for manufacturing?
“For us, the investment in creating standard SEMI or SECS GEM communication protocol is paying dividends when it gets to deployment of new technologies based on IIoT.
“The communication with all equipment types using the same protocols, and the identical user interface [that] technicians use in all manufacturing areas, makes any solution easily transferable and deployed across all factories. This equipment standard is the baseline that allows for the automation to be built-on.
“Beside the standards, the ability to acquire, clean, contextualize, store, process, merge, reason and act upon a growing amount of data is what makes the difference. At Intel we collect five billion data points per factory per day, and the demand is growing. Technology provides new ways to handle the data and derive meaningful actions based on it.”
How do you characterize smart-manufacturing use cases? What types are there?
“This is broad question, but it would be anything that automates any step of the decision-making based on real data. The factory becomes smarter as those decision-making loops, based on broader data, take actions in more autonomous way. To name just a few examples…
- “automatically adjusting the maintenance schedule based on equipment and process parameters;
- “using cameras and computer visions to detect defects and adjust or stop tools;
- “routing the material in line based on actual tool performance;
- “automated scheduling of material for the factory;
- “robotic material transport;
- “real-time interruption of the process on potential misprocess conditions;
- “providing the right data and context to streamline the decision making process for process and yield engineers.”
How will 5G impact manufacturing? How soon?
“We are expecting 5G to only impact areas of manufacturing that would typically be difficult to get infrastructure to – such as wide open spaces, outdoors or logistics.
“It is not yet clear if 5G will expand directly into a manufacturing floor that is already easy to get infrastructure to. That may primarily become an edge infrastructure cost-and-security discussion.”
How easy is it to make a costed business case for IIoT technologies?
“We use NPV (net present value) to measure a project’s ROI. We have seen projects in the past which were too expensive, which are now moving to positive ROI. In other cases the growing complexity of manufacturing forces us to rely on data to make decisions that could have been made by humans but are now becoming too complicated.
“The other side of this coin is production process enabling. If the automation/IoT work is needed to make the process work, it gets completed. Other items that impact cost, speed, and yield are ‘NPV’d’ and sorted by value.
“In addition to the above factors, I would also add driving increased tool availability and reducing human interaction.”
How important is collaboration (“co-creation”) to your working methods – with tech / solution vendors?
“For Intel, working with our equipment providers is key to ensuring equipment with advanced data collection is transferred to Intel systems properly. We also work with a few framework providers such as GE. But due to our size Intel is unique in the amount of proprietary, customized systems.
“Also, working with our semiconductor equipment suppliers to properly implement the SEMI standard is critical to our success.”
What are the biggest challenges to IIoT in manufacturing?
“The factories are running at high volume and IIoT projects take time and in many cases several iterations to show the return.”
How have you overcome this?
“We created a central technical group. This group assess every project, and develops methods, standards and ROI measurement. The group also provides technical support to help us keep improving, and offers integrative growing experiences to new project teams.
“This helps reduce project costs as we establish common IT services and infrastructure that intend to serve current and future demand.”
On August 22, 2018, Enterprise IoT Insights will publish a report and host a webinar on the role of digital technologies in industrial transformation in manufacuring. Go here to download the report after August 22. Go here to register for / listen to the webinar, with experts from ABI Research, ADLINK, Convergio, Hitachi Vantara, MESA International and PTC discussing seminal use cases and best practices that are setting the digital agenda for the industrial sector.
