STEWART CHALMERS and TOM HO, BISTel, Santa Clara, CA
Advances in Artificial intelligence (AI), cloud, big data, edge computing and internet of things (IoT) are pioneering the first AI knowledge sharing fab models. An array of AI smart manufacturing applications is fueling a resurgent semiconductor manufacturing industry. Smart applications are coming together to form the first knowledge sharing fab models. Understanding the power of the first AI fab knowledge sharing models is not enough, one must also understand the specific challenges.
Sharing the knowledge, means understanding the problems
How do you share the knowledge among all of the engineers in the factory? One of the challenges to knowledge sharing across the fab is complacent disengagement (FIGURE 1). This behavior underscores the difference between experienced engineers and inexperienced engineers. Experienced engineers bring the advantage of industry knowledge and “boots on ground” skills. It is advantageous for companies to share that knowledge across all the engineers in the fab. The fab of the future seeks to avoid siloed knowledge. Engagement and collaboration across the fab should be encouraged. This is ever more important given the aging of the semiconductor workforce. As the population ages, we have seen the difficulties our industry has had in attracting new talent. A dilemma is created. How do you shorten the gap between experience and inexperience? Knowledge is needed to accelerate response times solving problems quickly and efficiently.
Capturing knowledge and sharing it quickly among engineers
When engineers are empowered, smarter decisions are made, redundant work is eliminated, and problems can be targeted with efficiency (FIGURE 2). It is critical to achieving this objective that engineers have the tools to share knowledge across the organization.
What decisions are being made to solve certain problems?
In FIGURE 3, everyone has data and when we connect the data it becomes information and then knowledge. As we increase our knowledge, engineers began to enjoy greater insights. Engineers can then solve problems. Solving problems requires knowledge, experience and wisdom. It is also important that there is a path to getting there – the data evaluation process.
Knowledge sharing fab models empower engineers
Recent advances in AI smart manufacturing applications have created enabling technologies that have moved the knowledge sharing ball forward. AI, cloud, big data, edge computing and IoT technologies help capture and retain enormous data. These technologies empower engineers to make quicker and more insightful decisions. This is the bridge to efficiency. Today, we have the ability to gather much more information than ever, helping the modern fab to digest and transform data into knowledge and actionable insights.
Current smart manufacturing operates under the premise that decisions follow decision independent processes of detection, analysis and prediction. Detection identifies problems and alarms engineers that something is wrong. The second aspect of the process is analysis to pinpoint the root cause of the problem. Prediction involves identifying issues before they occur. It may also involves predicting the remaining useful life of critical equipment. These types of applications are now starting to incorporate AI elements that make the process quicker and smarter (FIGURE 4). Smart applications make analysis accurate and more exhaustive. By integrating AI into BISTel’s smart applications, BISTel enables engineers to be more efficient in identifying the problems. Understanding these issues helps us to create solutions that can predict the behavior of certain tools and processes. Many of today’s AI applications are mostly point of use. This is an evolution and, in the future, fabs will adopt a truly comprehensive knowledge sharing fab model where all data and insights are shared and understood.
How can we create a system where the system learns by itself?
How can we create a system where the system learns by itself? When it learns, the new Decision Support System (DSS) it offers diagnoses and then self-optimizes and, if warranted, self -corrects. In 2020, BISTel will introduce the industry’s first AI central platform for knowledge sharing across the fab and across the manufacturing ecosystem (FIGURE 5). The new Decision Support System (DSS) is smart enough to capture the learning but also act on the learning. This is what we refer to as the “knowledge sharing fab model”. Detection, analyses, and prediction integrated into one AI system that continues to build and share knowledge for the modern fab. The Decision Support System (DSS), conceptualized as a brain of sorts, discovers and captures all of the constant fragments of data extrapolated from the fab environment. This is a hallmark of the fab of the future.
AI DSS goes beyond the factory
The technology that provides for the decision support system could very easily reach beyond the factory (FIGURE 6). All have the potential to be revolutionized by such an autonomous system. It discovers new things and learns the relationship between cause and effect, and it uses that knowledge and acts on it independently. DSS has the power to connect across factories, organizations and industries to include supply chains, finance, IT, logistics and others. This is not just an application that resides within fab or factory. DSS can share knowledge across many different types of manufacturing organizations.
So, what do the first knowledge sharing fab models look like?
In the first knowledge sharing fab model, engineers are presented each morning with a personalized dashboard at their workspace. The dashboard may offer an update on the current status of the fab. Examples of updates may include illustrations representing number of units shipped, cycle time and yield status. Key activities of DSS may be depicted on an engineer’s dashboard. In FIGURE 7, key activities being performed by the DSS are presented – Activity, Recommendations, Reviews and Discovery.
The ACTIVITY feature provides information to the engineer on the health status of equipment and processes. In predictive based maintenance, knowledge sharing allows for an immediate influence on the fab and sub fab operations. For instance, in Fig. 7, if the DSS identifies a malfunction in a specific pump. The problem is highlighted on the engineer’s dashboard. This problem could be affecting the tool performance and yield in the fab. The system predicts the remaining useful life (RUL) of the asset and then creates a work order flow back to the enterprise resource planning system (ERP) where the system can order parts and/or schedule maintenance. This prediction also includes actions and recommendations for the engineer. The ability to predict when an asset will fail is having a huge impact on pharmaceutical, semiconductor and oil and gas manufacturing operations, reducing system downtime and saving considerable maintenance costs. The system logs the activity, as well as the correction, displays the incident and outcome and learns from it.
The RECOMMENDATIONS feature identifies problems and generates a set of solutions that is then presented to the engineer. Some solutions may call for approval. All of the work is done by the system and the human makes the decisions based on this knowledge capture.
A third feature Is REVIEW. If a novel sequence of events is presented, The DSS flags the engineer or operator for specific input to the system. If, for example, the system identifies an identical alarm repeat (IAR). The system will recognize the novel pattern, analysize the data and then, if needed, advises a review. The engineer may provide domain input to the system. Any subsequent input may be incorporated into a new protocol. In future, DSS would then identify and act autonomously. The knowledge has been captured and shared.
Feature four is DISCOVERY. This is an example of knowledge empowerment. If, in practice, the DSS, notes a trace alarm, records when it happened and on what lot, wafer and recipe and tool, DSS will then enrich the knowledge base. The system learns by itself. It transforms observed behavior from humans and processes, “connects the dots” and created a solution (FIGURE 8). DSS helps immensely by supporting quicker decision making and/or support.
AI based decision system support systems will proliferate over the next several years. New AI knowledge sharing fab models will play a leading role in shaping the fab of the future. Autonomous thinking accelerates our path to the fab of the future.
About the authors
Stewart Chalmers is VP Business Development and Marketing BISTel. He can be reached at stewart.chalmers@bistel.com Tom Ho is President & GM, BISTel America.
Stewart Chalmers Tom Ho
