Process Automation in Semiconductor Manufacturing: Issues and Solutions



Similar documents
Business-Driven Software Engineering Lecture 3 Foundations of Processes

Object-Oriented Systems Analysis and Design

An Automated Workflow System Geared Towards Consumer Goods and Services Companies

(BA122) Software Engineer s Workshop (SEW)

Journal of Information Technology Management SIGNS OF IT SOLUTIONS FAILURE: REASONS AND A PROPOSED SOLUTION ABSTRACT

Enterprise Integration: operational models of business processes and workflow systems *

A Symptom Extraction and Classification Method for Self-Management

INTELLIGENT DEFECT ANALYSIS, FRAMEWORK FOR INTEGRATED DATA MANAGEMENT

Collaborative & Integrated Network & Systems Management: Management Using Grid Technologies

Patterns in. Lecture 2 GoF Design Patterns Creational. Sharif University of Technology. Department of Computer Engineering

cesses relate human tasks that are rooted in the physical world. Such tasks include, moving, storing, transforming, measuring, and assembling

Chapter 2. Data Model. Database Systems: Design, Implementation, and Management, Sixth Edition, Rob and Coronel

A process-driven methodological approach for the design of telecommunications management systems

Enable Location-based Services with a Tracking Framework

An Integrated Methodology for Implementing ERP Systems

7. Classification. Business value. Structuring (repetition) Automation. Classification (after Leymann/Roller) Automation.

System Development and Life-Cycle Management (SDLCM) Methodology. Approval CISSCO Program Director

Automating Non-Standard Recipes In a Dual Gate Oxide Pre-Clean Process

Proceedings of the 6th Educators Symposium: Software Modeling in Education at MODELS 2010 (EduSymp 2010)

How To Develop Software

1 File Processing Systems

Run-time Variability Issues in Software Product Lines

COURSE OUTLINE. Track 1 Advanced Data Modeling, Analysis and Design

COMPUTER INTEGRATED MANUFACTURING

Tool Support for Software Variability Management and Product Derivation in Software Product Lines

A Faster Way to Temporarily Redirect the Role Based Access Control Workflow Processes Christine Liang

i-questionnaire A Software Service Tool for Data

Applying the Chronographical Approach to the Modelling of Multistorey Building Projects

Software Engineering. Software Processes. Based on Software Engineering, 7 th Edition by Ian Sommerville

Distributed Database for Environmental Data Integration

Business Process Modeling and Standardization

How To Create An Enterprise Class Model Driven Integration

BPMN PATTERNS USED IN MANAGEMENT INFORMATION SYSTEMS

MDE Adoption in Industry: Challenges and Success Criteria

DATABASE MANAGEMENT SYSTEMS IN ENGINEERING

Sales and Operations Planning in Company Supply Chain Based on Heuristics and Data Warehousing Technology

Taxonomy Development Business Classification Schemes Charmaine M. Brooks, CRM

Curriculum Map. Discipline: Computer Science Course: C++

Table of Contents. CHAPTER 1 Web-Based Systems 1. CHAPTER 2 Web Engineering 12. CHAPTER 3 A Web Engineering Process 24

Umbrella: A New Component-Based Software Development Model

Report on the Dagstuhl Seminar Data Quality on the Web

General Problem Solving Model. Software Development Methodology. Chapter 2A

Contents. Introduction and System Engineering 1. Introduction 2. Software Process and Methodology 16. System Engineering 53

Applying 4+1 View Architecture with UML 2. White Paper

2. MOTIVATING SCENARIOS 1. INTRODUCTION

Basic Trends of Modern Software Development

Software, Process, Business Process and Software Process

DATA QUALITY MATURITY

To meet the requirements of demanding new

ASCETiC Whitepaper. Motivation. ASCETiC Toolbox Business Goals. Approach

The Power of Two: Combining Lean Six Sigma and BPM

Faculty of Engineering and Science Curriculum - Aalborg University

Analyzing the Scope of a Change in a Business Process Model

An Object Model for Business Applications

CHAPTER 1. Introduction to CAD/CAM/CAE Systems

The Phases of an Object-Oriented Application

Demystified CONTENTS Acknowledgments xvii Introduction xix CHAPTER 1 Database Fundamentals CHAPTER 2 Exploring Relational Database Components

Managing a Fibre Channel Storage Area Network

To introduce software process models To describe three generic process models and when they may be used

Software Engineering. What is a system?

The Integration of Agent Technology and Data Warehouse into Executive Banking Information System (EBIS) Architecture

Building Systems Using Analysis Patterns Eduardo B. Fernandez Florida Atlantic University Boca Raton, FL

BPMN by example. Bizagi Suite. Copyright 2014 Bizagi

Complex Information Management Using a Framework Supported by ECA Rules in XML

A UML 2 Profile for Business Process Modelling *

Methodology of performance evaluation of integrated service systems with timeout control scheme

Information Systems and Technologies in Organizations

COCOVILA Compiler-Compiler for Visual Languages

IndustrialIT System 800xA Engineering

Enterprise Integration Architectures for the Financial Services and Insurance Industries

IDENTIFYING PATTERNS OF WORKFLOW DESIGN RELYING ON ORGANIZATIONAL STRUCTURE ASPECTS

IMPROVING PRODUCTIVITY USING STANDARD MATHEMATICAL PROGRAMMING SOFTWARE

A Complete Model of the Supermarket Business

Decomposition into Parts. Software Engineering, Lecture 4. Data and Function Cohesion. Allocation of Functions and Data. Component Interfaces

Introduction to Software Performance Engineering

Semantic Business Process Management Lectuer 1 - Introduction

AMFIBIA: A Meta-Model for the Integration of Business Process Modelling Aspects

State-of-Art (SoA) System-on-Chip (SoC) Design HPC SoC Workshop

MODEL DRIVEN DEVELOPMENT OF BUSINESS PROCESS MONITORING AND CONTROL SYSTEMS

Development/Maintenance/Reuse: Software Evolution in Product Lines

The Evolution Of Prototype Architectures Developed For The Scheduling Software Integration Project

Business Process Management Using. BPM Using Process Algebra and Relational Database Model

Copyright 2007 Ramez Elmasri and Shamkant B. Navathe. Slide 29-1

Component visualization methods for large legacy software in C/C++

MULTI AGENT-BASED DISTRIBUTED DATA MINING

BIG DATA: BIG CHALLENGE FOR SOFTWARE TESTERS

Integration of Time Management in the Digital Factory

Training Management System for Aircraft Engineering: indexing and retrieval of Corporate Learning Object

The «SQALE» Analysis Model An analysis model compliant with the representation condition for assessing the Quality of Software Source Code

Data Analysis 1. SET08104 Database Systems. Napier University

Software maintenance. Software Maintenance. Fundamental problems. Maintenance activities. Lehman/Belady model of evolution. Topics

what operations can it perform? how does it perform them? on what kind of data? where are instructions and data stored?

NASCIO EA Development Tool-Kit Solution Architecture. Version 3.0

A Multidatabase System as 4-Tiered Client-Server Distributed Heterogeneous Database System

ECS 165A: Introduction to Database Systems

JOB DESCRIPTION APPLICATION LEAD

LASTLINE WHITEPAPER. In-Depth Analysis of Malware

BUSINESS RULES AS PART OF INFORMATION SYSTEMS LIFE CYCLE: POSSIBLE SCENARIOS Kestutis Kapocius 1,2,3, Gintautas Garsva 1,2,4

Business Modeling with UML

Feasibility Study into the use of Service Oriented Architecture within the Atlantis University Portal

Transcription:

Process Automation in Semiconductor Manufacturing: Issues and Solutions Nauman Chaudhry, James Moyne, and Elke A. Rundensteiner Department of Electrical Engineering and Computer Science University of Michigan, Ann Arbor, MI 48109-2122 email: {chaudhry, moyne, rundenst}@eecs.umich.edu Abstract: While the area of process automation in semiconductor manufacturing has many issues in common with process automation and enactment in business processes, the nature of the semiconductor manufacturing process introduces a number of challenging new issues. Among these issues is the need to not only automate the process but also to continuously control and optimize it. Furthermore the semantics of success/failure for a semiconductor manufacturing process are quite different from the typical binary success/failure of business processes. In this paper, the issues in process automation in semiconductor manufacturing are presented, and compared and contrasted with the issues in traditional workflow technology. We then describe a system, based on an active database management system, that we have developed for process automation in semiconductor manufacturing. We also discuss the approach of this system towards addressing the different issues in the automation of semiconductor manufacturing processes. Keywords: Workflow automation, process automation, active databases, semiconductor manufacturing automation and control. 1 Characterizing the Semiconductor Manufacturing Enterprise Workflow technology is primarily aimed towards modeling, re-engineering and automating business processes. Workflow management systems have been characterized along different dimensions. When comparing workflow management systems with manufacturing systems, the characterization of workflow management systems along the dimensions of repetitiveness, predictability, initiation and control of the workflow is of particular interest. Along this dimension, the workflows consisting of repetitive, complex business processes with interaction of the information system with the business process and use of automated task performers are classified as production workflows. A process in workflow systems is defined as a coordinated set of process activities that are connected in order to achieve a common goal. A process in general can be composed of a hierarchy of sub-processes (or steps) along with different types of dependencies between these sub-processes. The workflow system then provides capabilities to model and specify, as well as to implement and automate the process [KrS95]. A semiconductor manufacturing enterprise when considered at the level of the business process is, of course, very similar to other business processes. However, a look at the requirements for process automation at the level of the semiconductor manufacturing facility reveals that notwithstanding a number of commonalities between automating business processes and semiconductor manufacturing processes, several interesting issues appear in semiconductor manufacturing that do not arise in business processes. These various issues are discussed in Section 2. This is followed in Section 3 by a description of the system being developed at the University of Michigan to address these issues, while Section 4 contains some conclusions. 1

2 Issues in Semiconductor Manufacturing Automation and Control Semiconductor wafer fabrication is a very complex manufacturing process. Production of each wafer may require several hundred individual activities (or steps as they are referred to in semiconductor manufacturing systems). The computer integrated manufacturing (CIM) system of a semiconductor manufacturing facility needs to provide facilities for factory management, planning of factory operation, scheduling of factory resources, specification and manipulation of processes, monitoring overall factory performance, and monitoring and control of the equipment in the facility. The CIM system may be conceptually divided into the following three layers [BCD94] (see Figure 1): The Corporate Layer serves as an interface between the manufacturing facility and (the workflow system of) the business process. Decisions about the quantity and type of wafers to be fabricated are communicated to the facility using this layer. The interaction of applications at this layer is quite predictable and can be automated using traditional workflow techniques. The Factory Layer contains applications that make decisions about when and where each wafer is to be fabricated, and what sequence of steps is required for this fabrication. Applications at this layer interact with the corporate layer and translate the decisions about the quantity and type of wafers to fabricate into an executable plan. Applications at this layer coordinate the resources of the factory to execute the fabrication process, checking dependencies and carrying out control. The Process Layer carries out the actual execution of each individual activity at different pieces of equipment. Applications (usually specific to each piece of equipment) communicate with the factory layer and carry out the process as specified by the factory layer. As may be gathered from the above outline, applications in the factory layer perform the main function of specification, coordination and execution of CIM processes analogous to similar activities in business workflow systems. A process specification application, called the Process Flow Specification Manager or PFM, is used to specify the fabrication process. The fabrication processes are specified in terms of a hierarchy of other processes which in turn may be formed of other processes and steps. A Scheduler keeps track of the equipment in the factory and based on this information, and the process specification, directs the process layer to carry out the individual steps on a wafer at a particular equipment, after which the wafer is transferred to the next equipment (Figure 1). In terms of the hierarchical specification of the process and the need to enforce inter-task dependencies, process automation in semiconductor manufacturing appears very similar to the corresponding functionality in production workflow systems. However, as we discuss below, automation of semiconductor manufacturing process also requires dealing with a number of new issues including the following. Tuning the process: A major activity that a semiconductor automation framework has to provide is the control of the fabrication process as a wafer undergoes a variety of fabrication steps. This requirement is caused by the fact that the processes often drift and shift with time due to a variety of factors such as equipment aging, variability of consumables, and fluctuation in ambient conditions. Performance may also change drastically after maintenance operations. The CIM system needs to keep track of the state of a wafer as it proceeds in the facility and carry out appropriate actions to maintain the overall quality of the fabrication. The requirement from the automation system, thus, is not only to specify and enact the process but also to make provisions for continuous tuning of this process depending upon the conditions. A related side issue is the support 2

Factory Manager Corporate Control Layer PFM Scheduler Process control advice Factory Control Layer Active Controller Rules for multistep control Process result Process Control Layer Equipment 1... Equipment n FIGURE 1. A simplified diagram of applications in a semiconductor manufacturing facility. for the dynamic evolution of the process. As the control actions are carried out, the process specification dynamically evolves in accordance with the control actions. The automation and control system should provide support for this dynamic evolution of the process. Definition of statistical control rules over continuous variables: Unlike the binary success/failure results of the activities in a typical business processes (e.g., transaction succeeds, claim rejected), the results of steps in a semiconductor manufacturing process are defined over continuous variables (e.g., uniformity, etch depth, etc.). Additionally, in many cases the control actions are taken based on statistical and/or imprecise estimates of these variables. Thus the process automation system should support the definition of control rules over statistical parameters and the definition of imprecise control rules. 3 Solution for Process Automation and Control: The Active Controller The AC Architecture: To provide a framework for monitoring and control of a semiconductor manufacturing process over multiple processing steps, we have developed a software controller, the Active Controller (AC) [CMR95]. This controller is built utilizing active database technology. Multi-step control is facilitated through the use of Event-Condition-Action (ECA) rules of the database. In a simplified form, these rules are utilized for multi-step control as follows: execution of a process step at an equipment or a composition of many such process steps constitutes the (composite) event for the AC rules. The conditions of these rules are the scenarios in which multi-step control can be used to improve the processing of the product, i.e., the situations in which processing at certain steps can be adjusted to compensate for errors in processing at other steps. These 3

conditions can be defined in terms of relevant process parameters, such as predicted process mean, specification limits, etc., and parameters of the manufacturing process, e.g., acceptable yield loss, etc. Firing of the rules causes appropriate actions to be taken to compensate for the error in processing. As the process steps at the different equipments are completed, the AC is notified of the results of these process steps. The AC updates its active database with this information. The AC active database detects if the completion of any steps causes an event of interest to happen. If this is the case, depending upon the result of this process step and relevant previous result data, multi-step control rules in the AC may be fired. A prototype of the AC has been developed at the University of Michigan using the ODE active database management system (DBMS) from AT&T. We are also building a test-bed to use this controller in a simulation environment as well as on actual processing equipment at the Display Technology & Manufacturing Center at the University of Michigan. Tuning the process: In current semiconductor manufacturing CIM systems the process is generally specified in the PFM by a hierarchical object-oriented data structure. We use ODE to specify the multi-step control knowledge in terms of ECA rules defined on the process data structure. When the rules fire, this process data structure is modified, thus changing the processing to be carried out on the wafer. The simpler modifications consist of tuning the process by changing just the parameters associated with one or more steps (e.g., modifying the depth of the etch step to compensate for the variation from the target in the execution of the deposition step). Such modifications can be easily accommodated by changing the appropriate attributes of the relevant step objects. Evolving the process structure: The AC control actions can also suggest changes to the process structure itself. Examples include a control advise to repeat certain steps in the process (e.g., adding a rework loop to repeat a set of steps already carried out) or an advise to add/delete steps in the process (e.g., replacing a sequence of automated steps by activities to be performed by a human operator). Although, in certain cases these changes can be accommodated by modifying only the process instances (e.g., by switching between different step instances of the same class), in other cases these suggested changes would require evolving the structure of the process (e.g., a suggestion to use a new sequence of steps which cannot be accommodated within the existing schema or a control advise to add certain steps to a process which require monitoring of a different set of variables). This evolution is handled by considering the original definition of the process as a prototype which is modified by the control actions. The evolution of the process structure also requires evolution of the rules defined on this structure. For example, new rules must be created to operate on the modified process class definition or under modified parameter settings. While this can sometimes be achieved via rule inheritance, in other cases it may require both rule and process schema evolution. We plan to further investigate this issue of rule evolution. Definition of statistical control rules over continuous variables: The results of steps in a semiconductor manufacturing process are defined over continuous variables (e.g., uniformity, etch depth, etc.). Certain control actions are defined based on SPC (statistical process control) rules defined over these parameters. To enact such rules, the AC needs to monitor the history of a parameter s value and to compare this value to statistical charts. For now, we are handling these rules by using ODE s facility for complex event detection (thus track- 4

ing the history of a parameter) and the ability to call functions in the condition part (thereby hiding the statistical control conditions in the code). However, a more elegant solution should allow direct definition of statistical control rules. We are thus investigating the extension of the rule language to allow for the direct definition of statistical control rules. Definition of imprecise rules: The lack of precise models for many semiconductor processes means that it is often difficult to interpret precise results of a process into precise actions to be taken to improve processing. The multi-step control rules thus may be imprecise. To provide this capability, we are currently focussing our attention on the issues of modeling and handling imprecise data and rules in active object-oriented databases. Note that for this research effort, we are building upon our previous research, described in [CMR94], carried out to incorporate imprecision in the relational control database used by equipment controllers in the process layer (Figure 1). 4 Conclusions In this paper we have described the AC, a software controller that has been developed to automate as well as control semiconductor manufacturing processes. The AC tunes the process by evolving the process definition as a result of control action taken in response to the execution of process steps. We note that the process tuning aspect of the AC can be further enhanced by utilizing accumulated process data and available rules to learn new rules. Although our present effort is directed towards semiconductor manufacturing, the same concepts can also be translated to other manufacturing environments. In addition to the workflow concepts considered in this paper, other workflow concepts are also relevant to semiconductor manufacturing and need further investigation. These include utilizing techniques used for agent selection in workflow systems for the selection of equipment in the CIM Scheduler and considering a distributed architecture for the AC. References [BCD94] R. Beaver, A. Coleman, D. Draheim, and A. Hoffman, Architecture and Overview of MMST Machine Control, IEEE Transactions on Semiconductor Manufacturing, 7, 2 (May 1994), 127-133. [CMR94] Nauman A. Chaudhry, James R. Moyne, Elke A. Rundensteiner, A Design Methodology for Databases with Uncertain Data, 7th International Working Conference on Scientific and Statistical Database Management, Charlottesville, VA, September 28-30, 1994, 32-41. [CMR95] Nauman A. Chaudhry, James R. Moyne, Elke A. Rundensteiner, A Generic Framework for Inter- Cell Control of a Semiconductor Manufacturing Facility, 42nd National Symposium of the American Vacuum Society, Minneapolis, MN, October 16-20, 1995. [KrS95] N. Krishnakumar, A. Sheth, Managing Heterogeneous Multi-system Tasks to Support Enterprisewide Operations, Distributed and Parallel Databases, 1995, 1-33. 5