Mani Janakiram

Factory cycle-time prediction with a data-mining approach

An estimate of cycle time for a product in a factory is critical to semiconductor manufacturers (and in other industries) to assess customer due dates, schedule resources and actions for anticipated job completions, and to monitor the operation. Historical data can be used to learn a predictive model for cycle time based on measured and calculated process metrics (such as work-in-progress at specific operations, lot priority, product type, and so forth). Such a method is relatively easy to develop and maintain. Modern data mining algorithms are used to develop nonlinear predictors applicable to the majority of process lots, and three methods are compared here. They are compared with respect to performance in actual manufacturing data (to predict times for both final and intermediate steps) and for the feasibility to maintain and rebuild the model.

Combining SPC and EPC in a Hybrid Industry

Integration of statistical process control (SPC) and engineering process control (EPC) is finding wide recognition and is successfully used in continuous process industries. However, application of this technique to parts or hybrid industries involving ..

Optimal Preventive Maintenance Scheduling in Semiconductor Manufacturing Systems: Software Tool and Simulation Case Studies

This paper presents the architecture and implementation of a preventive maintenance optimization software tool (PMOST), based on algorithms for the optimal scheduling of preventive maintenance (PM) tasks in semiconductor manufacturing operations. We also present results from four complex simulation case studies, based on real industrial data and employing full fab models, to illustrate the use, data needs and outcomes produced by PMOST. These results demonstrate significant improvements in tool production and consolidation of PM tasks. We give a description of the different software modules that compose PMOST, to provide guidelines as well as a template for other implementations of the PM optimization algorithms utilized by PMOST.

Model Context Selection for Run-to-Run Control

In the design of run-to-run controllers one is usually faced with the problem of selecting a model structure that best explains the variability in the data. The variable selection problem often becomes more complex when there are large numbers of candidate variables and the usual regression modeling assumptions are not satisfied. This paper proposes a model selection approach that uses ideas from the statistical linear models and stepwise regression literature to identify the context variables that contribute most to the autocorrelation and to the offsets in the data. A simulation example and an application to lithography alignment control are presented to illustrate the approach.

Operational modeling and simulation in semiconductor manufacturing

We present a panel session on the role of simulation in improving semiconductor fab operations. The participants include three principal investigators (PIs) from the recently awarded three-year,

Cycle time reduction at Motorola's ACT fab

1.2 million contracts sponsored jointly by the National Science Foundation (NSF) and the Semiconductor Research Corporation (SRC) on Operational Methods in Semiconductor Manufacturing; the Factory Sciences Program Director from SRC; and industry representatives from the semiconductor manufacturers and from discrete-event simulation vendors. Included are initial position statements from the various participants, which formed the basis for the panel discussions. For the industry participants, the statements may include, but were not limited to, specific important problems related to the role of simulation in operations in their respective companies, noting any significant technical, managerial, market, or other barriers. The position statements of the academic PIs describe the role that simulation is expected to play in their ongoing research in semiconductor manufacturing and/or their views on the key to successful application of simulation in the industry.

Real-Time Lithography Registration, Exposure, and Focus Control—A Framework for Success

This paper deals with the application of theory of constraints (TOC) and simulation to fab cycle time reduction at Motorolas Advanced Custom Technologies (ACT) fab. The cycle time crossfunctional team with the goal of transferring technology rapidly, keeping in mind customer satisfaction, underwent training in the basic theory of constraints, performed benchmarking, developed custom cycle time reports and developed techniques to measure theoretical cycle time and used multiples of theoretical cycle time as a basis to compare the performance of the wafer fabs actual cycle time (this metric is used in addition to the other fab metrics like, Turns/WIP for measuring the fab performance). Since ACT is a development fab, provision was made in the study, to include varied product mix, reentrant flows, optimized WIP, a typical hold in the process, etc. Engineering judgments coupled with simulation techniques were used to identify the constraints and these constraints were classified as physical and procedural and were treated accordingly by the team. The team evaluated each constraints and devised solutions based on TOC to reduce the fab cycle time.

Scenario analysis of technology products with an agent-based simulation and data mining framework

Intel factories are being challenged by an increasingly diverse product and technology mix. Semiconductor manufacturers have successfully leveraged advanced process control (APC) for overlay and critical dimension (CD) in the lithography module. Trends are continuously corrected through in-line metrology feedback, keeping the process on target. Applying an APC system suitable for litho control in high-volume manufacturing has resulted in significant improvement in registration capability and potential for better exposure focus and CD control.

Fault detection for batch monitoring and discrete wavelet transforms

A framework is presented to simulate and analyze the effect of multiple business scenarios on the adoption behavior of a group of technology products. Diffusion is viewed as an emergent phenomenon that results from the interaction of consumers. An agent-based model is used in which potential adopters of technology product are allowed to be influenced by their local interactions within the social network. Along with social influence, the effect of product features is important and we ascribe feature sensing attributes to the consumer agents along with sensitivities to social influence. The model encompasses utility theory and discrete choice models in the decision-making process for the consumers. We use expressive machine learning algorithms that can handle complex, nonlinear, and interactive effects to identify important inputs that contribute to the model and to graphically summarize their effects. We present a realistic case study that demonstrates the ability of this framework to model changes in market shares for a group of products in response to business scenarios such as new product introduction and product discontinuation under different pricing strategies. The models and other tools developed here are envisioned to be a part of a recommender system that provides insights into the effects of various business scenarios on shaping market shares of different product groups.

THE USE OF FMEA IN PROCESS QUALITY IMPROVEMENT

Batch operations are encountered in many industries and measurements are often recorded from automated sensors. It is important to determine whether an unknown batch is normal or unusual given a set of reference batches from normal operations. The measurements from a single batch can contain temporal readings that comprise a large time series. A discrete wavelet transformation (DWT) is applied to use the time and frequency localization of wavelets to extract features. A large number of coefficients can result and several methods to create summary features from the denoised coefficients obtained from DWT are compared. Also, a new summary feature incorporates information from denoised wavelet coefficients. The proposed study considers discrete wavelet-decompositions combined with principal component analyses to summarize batch characteristics. Results were validated on an industry data set. Copyright