Industry News, Trends and Technology, and Standards Updates

The amount of data generated in a modern wafer fab is astounding – a terabyte of data can be produced in mere moments. And while the technologies harnessed to build modern chips are at the absolute bleeding edge and have been compared to the sacred by some reviewers, the technology used to gather data and monitor factory equipment is often, surprisingly, not.

Ripping out and replacing shop floor applications is rarely done. And so, when a fab is built, its data acquisition and control architecture is generally as permanent as the bulk gas transports often seen hulking on the exterior. Modern software architecture has changed dramatically over the last 10 years. Yet the fab and its support systems are relatively static as compared to the process equipment needed to build the latest chips. The result of this imbalance and inertia is that fabs are full of equipment generating oceans of bits and bytes while the managers are struggling to stay afloat and man the oars. This problem only gets worse as management climbs the corporate ladder, or the mast, to extend the metaphor. The people in the corporate suites trying to understand and manage the realities going on in multiple fabs are absolutely inundated and clinging to debris in the deluge. No enterprise-level software system was ever designed to deal with this Noachian torrent of data. Out of this necessity fab managers have been employing data center technology and software architectures more commonly used to run social media sites and Amazon than to run factories.

On the 8th of August 2022, Google Search went down for approximately 34 minutes due to a poorly planned software update.  What is surprising about this is that it had been years since Search had suffered an outage like this, and more years hence. Google’s search page is the most heavily visited site on the internet and yet has one of the best availability metrics and performance history.  There are undoubtedly many reasons for this success, but one of the most acclaimed is Google’s application server: Kubernetes.  Named for the Greek word for helmsman or navigator, Kubernetes has become the standard application server for modern software architectures.  Software purists will undoubtedly object, and say that Kubernetes, or K8s, merely orchestrates the deployment of software, but this is no longer true.  K8s has become an ecosystem that hosts scores of other software products that do everything imaginable from compiling, testing, packaging, deploying and monitoring all the code required to keep a product like Google Search running.  

Incidentally, many of the most successful K8s ecosystem products continue the nautical theme with names like Docker, Armada, and Helm.  

In 2015, Google, in partnership with Docker and others, gifted the K8s technology to the open-source community at the Linux Foundation.  The Cloud Native Compute Foundation (CNCF) project was announced at that time with the goal to unify the rather fragmented containerized approach to software deployments. 

At PDF Solutions and the Cimetrix Connectivity Group, we saw this transformation happening and decided to get out in front of it.  We’d been training shop floor engineers to drink from data firehoses for decades, showing them how to tap into the torrents and pull out just the manageable streams they required and were equipped to handle.  Now, thanks to the CNCF, we can build software that handles far more data and still survives in the 24x7 environments that wafer fab production demands. Cimetrix’s Sapience Manufacturing Hub is our answer to this. 

The Hub solves the problem of getting actionable shop floor data to the top floor for the semiconductor industry, akin to navigating between Scylla and Charybdis. Because the most complex wafers take 3-4 months to process through a fab with the number of process steps in the thousands, even dealing with a single wafer lot of data has proved to be an enormous challenge.  Consequently, cost data associated with materials, labor and rework are allocated equally across all the products in the fab over many months. The result is that detailed data about how much a particular wafer costs or the profit margins of a product are buried in the averages.

The Hub is used to gather clean data from the factory floor and aggregate it where it makes sense. Milestone processes are used to report aggregated data by product, order, lot and other relevant factors so that costs can be accurately accounted for at the enterprise level. When the costs for energy, materials and testing go up while yields fall, knowing the details is important.  The Sapience Manufacturing Hub is the first cloud-native platform that can scale using common and well-known data center tools and provide this data in a way that is useful to the corporate suite of applications.

If you want to know more, please reach out to us by clicking the button below. We are a group of engineers committed to working on the biggest and most insoluble problems facing the electronics industry and really enjoy collaborating about all things semiconductor and data science.  

A little over 3 years ago, the Elsevier publishing company decided that the topic of Smart Manufacturing had achieved enough global breadth and momentum to warrant in-depth treatment in an edited set of articles. As they researched the domain, they realized that it split neatly into two categories of material: a general set of concepts that are applicable in any industry, and example implementations that are specific to a small set of related industries.

The editors then consulted with a number of industrial and academic subject matter experts to subdivide each category into a set of topics, and the Table of Contents for each of the two volumes was born. The results were first published in August 2020, and are available from Elsevier:

• Volume 1: Concepts and Methods (426 pages) 
  
  “Research efforts in the past ten years have led to considerable advances in the concepts and methods of smart manufacturing. Smart Manufacturing: Concepts and Methods puts these advances in perspective, showing how process industries can benefit from these new techniques. The book consolidates results developed by leading academic and industrial groups in the area, providing a systematic, comprehensive coverage of conceptual and methodological advances made to date.”
  
  
• Volume 2: Applications and Case Studies (528 pages) 
  
  “Research efforts in the past decade have led to considerable advances in the concepts and methods of smart manufacturing. Smart Manufacturing: Applications and Case Studies includes information about the key applications of these new methods, as well as practitioners’ accounts of real-life applications and case studies.”

One of the editors was Dr. Thomas F. Edgar, a long-time professor in the University of Texas Chemical Engineering department who specialized in process modeling, control, and optimization. He is also credited by many as being the “father of semiconductor APC" (advanced process control), since a number of his graduates ended up in the automation/process engineering department at AMD/Austin and transformed this domain from spreadsheets and “sneakernet” to a fab-wide, model-based process control system.

Needless to say, I was honored when he called to say that Elsevier wanted a case study on the semiconductor industry’s use of smart manufacturing, and to ask if I would write that section. I readily agreed, and provided the chapter titled “Smart Manufacturing in the Semiconductor Industry: An Evolving Nexus of Business Drivers, Technologies, and Standards.” The abstract for that chapter follows:

“The semiconductor industry embarked on its own “Smart Manufacturing” journey well over 30 years ago, long before the term was coined. The continuous productivity improvements that we now take for granted are essential for creating and building the devices that fuel our electronics-based global economy and maintaining commercial viability in a hypercompetitive industry. However, what we have learned in the process is that like many scientific endeavors, it is a journey without a destination. As new market opportunities are met with new device and system technologies in an ever-changing business environment, the list of manufacturing challenges is never complete.

This is where the global Smart Manufacturing initiative enters the picture. Although its key tenets are not specific to the semiconductor industry, the attention it drew to this topic triggered the formation of the SEMI Smart Manufacturing Community, which now provides a forum for thought leaders across the semiconductor manufacturing value chain to focus on these important challenges. To put this initiative in its proper perspective, this chapter explores the past, present, and future of Smart Manufacturing in the semiconductor industry.”

Regarding the Smart Manufacturing Initiative (now with chapters in several geographic regions), SEMI is the perfect host for such a group. Since its charter includes both trade and standardization activities for the industry worldwide, and its members develop the diverse technologies that comprise the electronics industry, SEMI forms an ideal backdrop for describing the evolution of semiconductor Smart Manufacturing. This is illustrated in the figure below.

Thinking back over my almost 50 years in the industry, I think the factor that contributed the most to the industry’s current success was the evolution of its collaboration culture. Whether driven by the need for industry efficiency, the fear of extinction, or the recognition of mutual interdependence, today’s global semiconductor industry enjoys a collaborative culture that is unequaled in other for-profit industries.

If the topic of Smart Manufacturing piques your interest, I invite you to visit the Elsevier websites shown above for more information. I have also distilled the semiconductor case study chapter into a set of slides that I will be happy to review/discuss with you.

We wish you the best on your company’s Smart Manufacturing journey – let us know how we can help!

We’re drowning in data but still thirsting for wisdom… This is a paraphrase of what one of the virology doctors interviewed at the beginning of the COVID-19 pandemic said about the challenges they faced in charting a course of action using the myriad volumes of data being collected around the world. The sentiment reminded me that in an entirely separate but no less complex domain: electronics manufacturers face similar challenges using data from the thousands of disparate sources in a factory to make the decisions that ultimately determine the viability of their enterprises.

To the extent that these two domains have a similar problem statement, they also share a set of solution technologies. Among them is Machine Learning (ML), which is a promising subset of Artificial Intelligence (AI). The essence of Machine Learning is using data collected from past events to develop and train a model that can make reliable predictions about the outcome of future, as-yet-unseen events.

Machine Learning technology is already more ubiquitous than you might imagine. For instance, if you’ve ever 1) ordered one of Amazon’s product recommendations based on your previous purchases and search history, 2) seen the names on the faces of people in your photo library (and updated the ones it got wrong), 3) rated your satisfaction with a Netflix movie you’ve just watched, or 4) interacted with one of the language translation apps on your smart phone, congratulations—you are now an integral part of the Machine Learning ecosystem of the suppliers of those everyday products.

Beyond these familiar examples, two factors have contributed to the rapid ascendance of ML as a viable production technology: the availability of massive amounts of data and affordable storage and processing power to analyze it. As a result, Machine Learning is now seen as a key enabling technology for Smart Manufacturing in multiple industries and at multiple positions along the value chain.

Documented use cases in semiconductor front-end factories include production scheduling and dispatching, process analysis and excursion prediction, equipment FDC (fault detection and classification), preventive maintenance, failure prediction, trace data analysis, and so on. Equipment suppliers are likewise looking for ways to add value to their products by offering ML-based capabilities as an option. Semiconductor back-end and SMT factories and their respective equipment suppliers are also actively evaluating ML technologies. These instances all have one thing in common: the need for quality equipment data… and lots of it.

Two of the most prevalent approaches to Machine Learning are 1) supervised learning, which uses labelled data samples to develop a model that predicts the labels for future samples; and 2) unsupervised learning, which can use unlabeled data samples to develop a model of expected behavior that can predict deviations from that behavior for future samples.

Both categories have direct applications in manufacturing and include a range of specific algorithms that may be suited to various problem types. The trick is knowing which technique(s) to apply in which situations, and once chosen, how to tune a particular technique for a given problem.

One category of public-domain algorithms that has shown significant promise for commercial applications has been labelled “Deep Learning,” and it refers to a variety of “Artificial Neural Net” approaches. These algorithms fall into the supervised learning category but can nevertheless be applied in situations where the labels are not known a priori (e.g., analyzing system log files to detect anomalous behavior).

A frequently-cited candidate for a supervised learning algorithm is predictive maintenance, which uses as input a high-dimensional (i.e., lots of parameters) trace vector of equipment parameters with a binary label (“good” or “failed”). With enough production history this data can be used to train and validate a failure prediction model that would improve on the “run to fail” or “just in case” maintenance strategies used today.

Likewise, a good candidate for unsupervised learning is anomaly detection using equipment log file data (could also include trace data) as input. Many equipment suppliers save this kind of data for “after the fact” analysis of problems that occur in manufacturing without a clear sense of what might be useful. Deep Learning algorithms are particularly adept at using high dimensional data to create complex mathematical models that represent “normal” behavior and can be used to flag “abnormal” behavior as it happens.

From an implementation standpoint, there is no “one size fits all” Machine Learning package or library that can address the full range of potential opportunities. Moreover, since Machine Learning is only a component (albeit a vital one) of an overall solution system, it is important to understand how all the pieces fit together when considering these technologies for a specific problem. Subsequent posts in this series will explore these topics in greater detail, so stay tuned for additional information about how to map these new technologies into your product and problem space. However, if your need is more urgent, click the button below to see how we can explore this exciting new area together.