Industrie 4.0 Vision Versus Semiconductor Manufacturing Status Quo (Part 1)

Industrie 4.0 semi 1

State of the art 300mm semiconductor Fabs are currently the closest ‘real life’ implementations of the Industrie 4.0 1 vision. As a result, some in the industry may ask themselves, “what is new for us here?” A closer look at both the future vision and the current status quo will reveal many opportunities to generate new business value.

I stick to the German spelling of ‘Industrie’ for two reasons – 1) because the initiative was started in Germany – back in 2012, and 2) because I’m German. The core Industrie 4.0 concepts include vertical and horizontal integration and End-2-End integration of engineering throughout the value chain. These are already fundamental features of most semiconductor manufacturing companies – the modern 300mm Fab is a giant Cyber-Physical-System (CPS), with its automated material handling systems (AMHS), automated stockers, integrated equipment and intelligent facility management systems.

Long before the Industrie 4.0 ideas were articulated, the semiconductor industry started moving towards this type of advanced holistic automation, motivated by the investment required to build and run a 300mm Fab. More than a decade ago, a new 300mm Fab cost far beyond 3.5 Billion USD, and now it may be up to twice that amount. Back then, high utilization and productivity were already identified as critical requirements to make the investment profitable. In today’s business environment, semiconductor manufacturers are forced to drive the level of automation and optimization even further, due to shortening product cycles, increased number of products running concurrently in the Fabs, and the continued falling of selling prices.

Let’s have a closer look at the status quo of 300mm semiconductor manufacturing today, as compared to the Industrie 4.0 vision. I will consider vertical integration in this first part of my blog. In the second part I will examine horizontal integration, “Lot size 1”, and the necessary improvements that will drive 300mm (& 450mm) Fabs even closer to the promises of Industrie 4.0.

Vertical integration:

Vertical integration covers all the five ISA 95 2 Levels, from Level 0 = the physical equipment with all its sensors and actuators, up to Level 4 with the enterprise financial and planning systems. Integrating systems and equipment from a large set of suppliers requires standardization of physical interfaces for material handling as well as software interfaces and protocols. In the semiconductor industry such standardization was and is driven from the industry association Semiconductor Equipment and Materials International (SEMI) 3.

For the Machine to Machine (M2M) communication SEMI Equipment Communication Standards/Generic Equipment Model (SECS/GEM) was established. The M2M equipment communication uses High-Speed SECS Message Services (HSMS) via TCP/IP which is also a target for Industrie 4.0 to overcome the barriers of proprietary automation protocols.

Semiconductor process and measurement equipment produce huge amount of data = Big Data. In order not to impact the equipment control information data flow from the sensor and measurement data a second dedicated interface for Engineering Data Acquisition the so called ‘interface A’ was introduced. The equipment data is consumed from different kinds of analytics applications. One group of such analytics solutions focus on product and process quality like standard Statistical Process Control (SPS), Run to Run (R2R) control, Advanced Process Control (APC) and multi variant analysis for Fault Detection and Classification (FDC).

The second group is focused on parts logistics, WIP area scheduling and preventive equipment maintenance. The Big Data challenges ahead for the semiconductor industries are identified from the International Technology Roadmap for Semiconductors (ITRS) as increasing sensor data frequency from the equipment from currently 10-100Hz to up to >1 kHz in near future. This massive amount of data create the need for new Big Data platforms to store the data and solutions for streams data mining.

Figure 1 ITRS 2013Figure 1 ITRS 2013 – projected increase of equipment availability and data output frequency 4

According to ITRS the level of analytics used today is insufficient to address semiconductor manufacturing requirements with respect to yield learning, throughput optimization and Overall Equipment Effectiveness (OEE). To address these challenges, predictive analytics are required. This new requirement has been called ‘Augmenting Reactive with Predictive’ (ARP).5 It will enable new capabilities, such as predictive maintenance to achieve target equipment availability from 95% to >96%.
Already in 1998 SEMATECH published the “Computer Integrated Manufacturing (CIM) Framework Version 2.0” 5 that defines the functional scope, interfaces and information model for CIM systems on higher Levels. As part of the vertical integration these systems sit on top of and utilize the M2M layer.

Figure 2 Automated Material Handling System (AMHS)

Figure 2 Automated Material Handling System (AMHS)

Modern chips require more than 1500 process steps with different sequences depending on the product being built. Each part is automatically sent to the next piece of equipment to perform the next process step, or to an automated stocker. Similar to the Industrie 4.0 proposal the product knows how it gets built and what equipment is available and capable of performing the next required process step. A minor difference is that the information itself is not stored on the product or the standardized carriers (=FOUP) RFID, but in the Manufacturing Execution System (MES).

Currently the decision to which equipment a part should be sent to is based on configured rules that might not always result in an optimum solution for the overall Fab throughput. To overcome this disadvantage, analytics solutions are implemented to ‘schedule’ the parts path through a section of the overall flow. Current available algorithms and computing capabilities are still not advanced enough to schedule the end-to-end production of thousands of parts each, with thousands of process steps using hundreds of pieces of equipment. Changing the current mostly centralized approach of dispatching toward a distributed approach as suggested by Industrie 4.0, where each part communicates with the equipment to decide where to go next, will need to get evaluated and piloted.

More advanced scheduling capabilities are also required to achieve the semiconductor manufacturing Environment, Safety, and Health (ESH) goals – which are very much in line with the ‘resource and energy efficiency’ goals from Industrie 4.0. For equipment to decide if it can switch itself into an energy efficient idle mode (e.g. by reducing vacuum pump speed, reducing heating and/or cooling etc.), the equipment needs to know the expected arrival time of the next batch of parts to be processed. To define the optimum time for a preventive maintenance task, the ‘ultimate’ predictive maintenance application will need to know the planned future processing time for each piece of equipment, as well as the production idle time windows for the equipment.


Conclusion goals and vision of Industrie 4.0 vertical integration  versus 300mm manufacturing status quo:

Table 1

How the semiconductor manufacturing reality compares to the Industrie 4.0 vision for horizontal integration and  “Lot size 1”? What improvements will  be required to drive 300mm (& 450mm) Fabs even closer to the promises of Industrie 4.0 ? I will address this questions in the second part.



1 ”Recommendations for implementing the strategic initiative INDUSTRIE 4.0”
4 Factory Integration 2013 tables
5 “Computer Integrated Manufacturing (CIM) Framework Specification Version 2.0”


Subject Matter Expert, Global Electronics Center of Competence at IBM

More stories

Blockchain Solutions: Sometimes it takes a village (i.e. a consortium)

I’m often the person who presents the basics of blockchain technology to people at electronics companies, which brings a small but enjoyable reward with it: I get to see the light come on in executives’ eyes when they realize that blockchain can solve a real-life problem for them. Our discussions almost always get more animated […]

Continue reading

From Visual Inspection to the Cognitive Future in Electronics Industry – IBM EGLF 2017

By looking back at 2015, no one could possibly deny that the Fourth Industrial Revolution (Industry 4.0) was the most talked-about topic for transformation in the Electronics industry. Tiny Smart sensors and actuators coupling with newly available IT technologies, such as Cloud, Analytics, Mobile, Social Media, and Security gave birth of Internet of Things (IoT). […]

Continue reading

Can Blockchain and Cognitive Analytics yield higher demand forecasting accuracy?

Written by: Jaime Marijuán Castro In the electronics industry nowadays, the average gross value of inventory (all types) as a percentage of revenue is 9,58 % (*), a pretty high ratio and proves that the cost of inventory is a matter of concern within the industry. High levels of inventory can be tackled in multiple ways, […]

Continue reading