The semiconductor industry segment is known for the most cutting edge manufacturing technology, highest levels of automation and exacting, precision manufacturing. Given this perception, one would assume most semiconductor manufacturers would be fully reaping the benefits of Industry 4.0.
The reality is far from it. The 2021 global chip shortage bears testimony to the way in which the industry’s perceived capabilities have fallen short on reaping the full potential of digital transformation and automation.
Accenture reports that in 2021, Apple and Samsung will be forced to delay the launches of their next smartphone models. The auto industry will lose upwards of $60 billion in revenue due to chip shortages. Surely, the challenges posed by the pandemic might be one of the reasons for the lag in chip manufacturing. The pandemic should in fact have been a driving force towards automation, higher efficiency and better results for the industry. So what is causing the lag?
To understand why the Semiconductor industry is struggling to meet burgeoning demands and to automate at a higher pace, we need to understand the issues which exist in a typical semiconductor manufacturing plant. What makes it difficult to achieve full automation? To start, we need to understand that ‘full automation’ is not a stage but a journey. It doesn’t begin with deploying automated transport systems, AGVs, stockers and robots. It starts at a higher, strategic level, with an analysis of existing issues, solving analog problems with digital solutions.
Typical challenges on a semiconductor shop floor which prevents automation
The reliance on MULEs and point solutions
One of the major challenges which semiconductor plants face is that they rely on legacy systems, such as old, highly customized and frequently undocumented MES applications. We refer to these as MULEs: Mature Undocumented Legacy Execution systems. MULEs are especially problematic when it comes to the ability to integrate with enterprise, SCADA/plant floor and IoT applications. They are also notorious for needing countless hours of coding for making simple adjustments to activities like workflow. This renders such applications unsustainable in the long run.
Point solutions serve various purposes, from controlling a particular piece of equipment, enabling a set of equipment transactions, maintaining quality data, to routing materials and scheduling customer orders.
Point solutions present two major challenges:
- Lack of scope in the overall scheme of things. It causes their inability to graduate to a broader application footprint or system framework without intense coding and API creation.
- Second and most important, is the reliance of existing, highly specialized personnel on these systems and their resistance to change. Familiarity and comfort with systems is not a reason to maintain the status quo. Moving to a modern MES brings benefits not only in new functionality and easy configuration, but better visibility and productivity.
Improving Material Management
The use of labels and paper travelers is extremely commonplace even in modern day fabs. Material, as it moves through the shop floor, may not be tracked at all times, due to a myriad of applications orchestrating the operation. This lack of traceability and absence of an overarching system to visualize, track, and execute the entire workflow renders blind spots throughout the operation.
The inefficient and unplanned management of material is manifested in different ways. Carts may go missing for days on end; orders may be delayed beyond the traditional 18 week lead time. It is often reported that material handling and traceability issues lead to higher scrap rates, especially when the material enters rework, exiting the standard workflow cycle.
Missing material leads to inefficiencies in manufacturing and longer term delays in customer order fulfilment. At its extreme, it contributes to the shortage of chips and products in the market. Reliance on manual procedures in material handling is a major challenge in achieving plant-wide automation.
Unfortunately, the solution is not as simple as putting automated transport systems, AGVs and stockers in place in order to automate the material movement and control the material flow. At the heart of this issue is the lack of preparation associated with the inclusion of automation. Training, refined business practices, new equipment and applications are cascade effects of including automation.
A focus on localized issues and lack of automation strategy across the value chain
One of the reasons why semiconductor manufacturers are struggling to automate is that automation is considered a shop floor-level deliverable. In actuality, it goes far beyond material processing. Full automation begins with a strategic mindset to examine the manufacturing processes. This is to understand what can be automated (manual activities, machine integrations, work order management, material movements) through the application of digital solutions.
An approach to solve localized problems, ignoring the enterprise and value chain wide-application architecture, integration and execution risks, may create gaps and performance plateaus. Focusing on a specific piece of equipment, like a labeler, to increase its accuracy and speed, solves the immediate issue. However, it may create a ‘ripple effect’ of imbalance, causing a larger queue in the packaging area to keep up with the increased throughput.
Another common example is adding automation to warehouse management. Overhead transport systems or AGVs cannot transport semiconductor FOUPS (Front Opening Unified Pod), wafer carriers, cassettes or other wafer handling solutions without an MES orchestrating their movement. A lack of integration with an overarching system like an MES means the material moves through the floors automatically, but it remains missing and unavailable when needed for processing at a particular operation/equipment/station, as there’s no systematic context or tracking for this movement.
Full Automation Begins on the Drawing Board
A holistic, people-centric approach will lead to better automation efforts, but it must be executed in a well-planned manner. It should be based on a strategy which examines the status-quo with all its complexity, followed by the replacement of manual processes/steps, point solutions and legacy systems.
Let’s understand how any semiconductor manufacturer can pursue full automation and expect a higher degree of success. Two major steps towards automation are:
1. Defining the current process while mapping the need for core applications. This is further examined and simplified down to individual activities and process functions.
2. Careful selection and deployment of the right MES.
The most critical aspect, which is often overlooked in an unsuccessful transformation is the people. People and their contribution to the process are an extremely important contribution towards achieving advanced automation.
The first step in pursuing full automation should always be the clear definition of the existing business process. What activities form the whole process and where does the complexity lie? No digital transformation and automation drive can be complete without clearly demarcating the existing status of people, process and equipment from an applications perspective and expected future deliverables.
Any semiconductor manufacturer planning to increase their level of automation should begin with performing an Application Portfolio Rationalization (APR) exercise. This process entails a clear definition and mapping of the entire value chain. It determines the core applications needed, followed by a functional mapping to assist in the new application selection, existing application retirement and subsequent implementation.
APR focuses on the activities which lead to creation of value and defining the process which helps deliver the value. Activities need to be mapped clearly. For manufacturing, it would be from the time material orders are sent to suppliers, until the time customer orders are shipped (and if needed beyond) on either side of the supply chain. Once the entire value chain is mapped, it can be determined what core applications are needed.
If the ISA-95 Model is taken as the point of reference, at the enterprise level 4 it may be determined that ERP, WMS and CRM are needed, followed by a level 3 MES and level 1 or 2 automation application like SCADA or IoT.
Once the core applications needed to form the automation backbone are determined, functional mapping is performed. The goal is to examine the relevance of the existing point solutions, which may very well be in the hundreds. This is perhaps the most challenging part of understanding the status quo. Owners and stakeholders using point solutions often overestimate their need and overvalue their functional abilities. Subject matter experts may be able to determine whether or not a point solution is truly needed, and worth the effort to integrate it to the core applications or retire the application.
MES drives factory automation in Semiconductor processing
MES applications reside at a very critical juncture in the core manufacturing application schema. They allow true value chain wide application-based integration, which is essential for end-to-end visibility and control. The visibility, oversight and control are major deliverables of full automation. They result in better quality, higher process yields and overall stabilization and repeatability in operations.
Factory Automation allows specific workflows to be triggered to correctly handle events, such as calculating and coordinating the next equipment step or destination of in-process or completed lots, and guiding the appropriate automated transport system.
MES forms the master data repository for both enterprise and shop floor data. It is this very data, if obtained in real-time, contextualized and relayed to automation hardware, such as AGVs, wafer stockers and Robots, that then delivers advanced levels of automation, always progressing and self-optimizing towards a theoretical ‘lights out’ fab.
MES eliminates the need for point solutions and MULEs. It houses all execution, orchestration and control functionalities needed for automated operations management. The application integrates with process equipment, IoT devices and also manages production routes and material flow. It is capable of providing visibility, thereby reconnecting the existing disconnect which leads to missing material and production loss. It is the application, which either controls the MHS/MCS or performs the material management as a function, by directly controlling labelling in semi-automated processes and transport systems in fully automated ones.
Integrate, standardize and automate with MES
The MES acts as the framework for business process reengineering (BPR), creating standard templates. This provides a degree of standardization in all manufacturing plants. It eliminates the need to depend on local or home-grown solutions to create a composite value chain-wide structure. It is not only automated, but delivers with higher predictability, quality and efficiency.
Selecting an industry-specific MES will ensure that your specific manufacturing and operations challenges are covered, with the right equipment integration, taxonomy, nomenclature and workflows embedded into the MES.
So, when we look at full automation for a semiconductor fab, we must include the people, processes and products. Legacy systems, manual processes, and disconnected workflows are all hurdles to overcome. By clearly defining the current process, mapping the core applications needed, eliminating unnecessary legacy and point solutions, and implementing the right MES can be your solution for full factory automation.