Virtual metrology for copper-clad laminate manufacturing

doi:10.1016/j.cie.2017.04.016

Computers & Industrial Engineering

Volume 109, July 2017, Pages 280-287

https://doi.org/10.1016/j.cie.2017.04.016 Get rights and content

Highlights

•
Virtual metrology(VM) applied to real data of Copper-clad laminate(CCL) manufacturing.
•
Through modeling tests, we found variables - easily ignored on work-sites - with significant possible impact on outcome.
•
Effective monitoring by focusing on important variables intensively is enabled.
•
By applying VM, it is possible to improve productivity by reducing time and cost.

Abstract

Copper-clad laminate (CCL), the key material for printed circuit board production, is used in various electronic products; thereby, the demand for CCL is on the rise. The process of CCL manufacturing occurs in three phases: treating, lay-up, and pressing, while the process with the largest influence on quality control is the treating. For effective quality control, the treating process requires intermediate inspection for three important quality factors: treated weight, minimum viscosity, and gel time. However, a manual inspection, which present-day manufacturers perform, incurs heavy cost in terms of time and money, rendering it ineffective. This study proposes the application of virtual metrology for CCL manufacturing to predict product quality derived from processing data without a product quality inspection. The actual process data from a CCL manufacturer in Korea was collected for a duration of approximately 5 months. Based on these data, the application builds a prediction model for CCL quality by utilizing the process variables affecting the CCL quality as predictor variables. As a result, four regression algorithms and three methods of variable selection were applied to build the prediction models for virtual metrology. Prediction models were obtained with a high accuracy in specific target variables. It was also verified that quality control was influenced by not only the important predictor variables empirically recognized by process engineers in the field but also by several essential variables previously unknown to the engineers; effective quality control will be possible by focusing on these variables particularly and more efficiently instead of overall monitoring.

Graphical abstract

Introduction

Copper-clad laminate (CCL) is the core element of a printed circuit board, which is used in a number of electronic products such as smartphones, tablets, digital cameras, and laptops; demand for CCL has increased significantly with rapid development in the information technology industry. CCL manufacturing is divided into three basic processes: treating, lay-up, and pressing, as shown in Fig. 1.

Treating is the process that plates glass cloth with resin by using immersion and rollers, and then produces prepregs supplying heat by air-circulating or by using infrared oven-drying (Khandpur, 2005). Lay-up is the process that assembles treated prepregs and copper foils for pressing. Copper foil is laid against a stainless steel press plate and a number of prepregs are stacked on top of the copper. The desired thickness of the laminate determines the number of layers, and the pressing process produces fully cured laminates by applying heat and pressure to packs simultaneously. The pressure is based on hydraulic operations, and steam is a typical heat source. The treating process is the most significant among these three processes. The treating process is performed under very strict supervision, and three metrology variables - treated weight, minimum viscosity, and gel time are inspected in the middle of the treating process. Currently, process engineers implement actual metrology for all inspections, but this causes decreasing productivity because a lot of time and cost is required for testing. Moreover, it is necessary to extract the key variables based on data for efficient monitoring.

The solutions for these problems can be considered as a virtual metrology estimate by using prediction models (Lin, Hung, Lin, & Cheng, 2006). Virtual metrology is the methodology to build prediction models with historical data, which is collected from production equipment sensors, so engineers inspect the product partially and predict the quality for the rest by using prediction models. Virtual metrology is to be applied to various manufacturing processes, typically semiconductor processes, and a detailed description of virtual metrology will be described in Section 3.1.

The purpose of this paper is to apply the virtual metrology system to CCL manufacturing, particularly three quality variables in the treating process: treated weight, minimum viscosity, and gel time. For learning prediction models, we used input variables consisting of process variables, material variables, and environmental variables, and three quality variables were used as target variables. The real-world data collected by a CCL manufacturer in Korea during approximately 5 months for two CCL products was applied to virtual metrology, by using four prediction models and three variable selection algorithms to build prediction models. As a result, we were able to derive important variables and predict target variables, and verified these variables through reviews with the process engineer.

The remainder of this paper is structured as follows: first, we will give a brief review of previous research on application of virtual metrology, and we will describe the virtual metrology system and the algorithms used in this research in Section 3. Then, in Section 4, we introduce the data overview, preprocessing, and experimental results. Finally, we propose future work with a conclusion.

Section snippets

Related work

The study on PCB - increasingly becoming more crucial to our lives as a key part of the electronic equipment - is actively being made on many researches (Chang et al., 2006a, Kim et al., 2013). A hybrid-GA algorithm is used in scheduling problem and component placement of the sequencing problem for PCB assembly, in order to reduce time and cost (Chang et al., 2007, Hardas et al., 2008). The study of CCL manufacturing is critical because CCL is one of the key components of PCB. Although CCL

Virtual metrology

In this paper, we apply virtual metrology to three metrology variables in the treating process, which is the core process of CCL manufacturing, and those variables are treated weight, minimum viscosity, and gel time. As stated above, virtual metrology is the method to build prediction models with previous data, which is collected from production equipment sensors, so engineers must investigate a part of the product and estimate the quality of the rest of the product by using prediction models.

Data description and preprocessing

We used real-world data collected from a CCL manufacturing company in the Republic of Korea. The data contained material information, the manufacturing environment, and process parameters in the treating process. Input variables include temperature, velocity, weight, pressure, tension measured by sensors as well as manufacturing variables obtained from opinion of domain experts. This work is to apply virtual metrology regarding two CCL products produced by the different equipment in CCL

Conclusion

This study built prediction models by applying the virtual metrology system for three quality factors of CCL manufacturing: treated weight, minimum viscosity, and gel time. We applied three different variable selection methods and four prediction models. In our experimental results, the SVR-GA outperformed for treated weight, and the RF-GA was found to be the best for minimum viscosity and gel time.

By applying virtual metrology to CCL manufacturing, sampling inspection will be performed instead

Acknowledgements

This work was supported by the BK21 Plus Program(Center for Sustainable and Innovative Industrial Systems, Dept. of Industrial Engineering, Seoul National University) funded by the Ministry of Education, Korea (No. 21A20130012638), the National Research Foundation (NRF) grant funded by the Korea government (MSIP) (No. 2011-0030814), and the Institute for Industrial Systems Innovation of SNU.

References (31)

M. Boin et al.
New methods of metrology data analysis during semiconductor processing and application to rapid thermal processing
Materials Science in Semiconductor Processing
(1998)
P.C. Chang et al.
Adaptive multi-objective genetic algorithms for scheduling of drilling operation in printed circuit board industry
Applied Soft Computing
(2007)
P.C. Chang et al.
A hybrid model by clustering and evolving fuzzy rules for sales decision supports in printed circuit board industry
Decision Support Systems
(2006)
C.S. Hardas et al.
Development of a genetic algorithm for component placement sequence optimization in printed circuit board assembly
Computers & Industrial Engineering
(2008)
K. Izui et al.
Multiobjective layout optimization of robotic cellular manufacturing systems
Computers & Industrial Engineering
(2013)
G. Kanagaraj et al.
A hybrid cuckoo search and genetic algorithm for reliability–redundancy allocation problems
Computers & Industrial Engineering
(2013)
P. Kang et al.
A virtual metrology system for semiconductor manufacturing
Expert Systems with Applications
(2009)
T.S. Li
Applying wavelets transform and support vector machine for copper clad laminate defects classification
Computers & Industrial Engineering
(2009)
H.C. Lu et al.
An integrated algorithm for cutting stock problems in the thin-film transistor liquid crystal display industry
Computers & Industrial Engineering
(2013)
R. Slavkovic et al.
An application of learning machine methods in prediction of wear rate of wear resistant casting parts
Computers & Industrial Engineering
(2013)

A.J. Su et al.

Control relevant issues in semiconductor manufacturing: Overview with some new results

Control Engineering Practice

(2007)

E. Yu et al.

Ensemble based on GA wrapper feature selection

Computers & Industrial Engineering

(2006)

J. Besnard et al.

Wafer-to-wafer virtual metrology applied to run-to-run control

L. Breiman

Random forests

Machine Learning

(2001)

B. Bunday et al.

Value-added metrology

IEEE Transactions on Semiconductor Manufacturing

(2007)

Cited by (16)

Virtual metrology in semiconductor manufacturing: Current status and future prospects
2024, Expert Systems with Applications
Advanced Process Control (APC) has become an increasingly pressing issue for the semiconductor industry, particularly in the new era of sub-5 nm process technology. To minimize human input in analyzing equipment state and output quality, Virtual Metrology (VM), one such incredibly reliable and resource-saving APC tool built on artificial intelligence and machine learning techniques, was introduced as an aide to real metrology and is aimed at predicting semiconductor product quality based on equipment status/process variables and sampled actual metrology. In this paper, VM efforts for semiconductor manufacturing applications are, for the first time, systematically reviewed. Based on comprehensive mining of literature, a generic architecture for semiconductor VM is presented, and the general requirements are discussed. Moreover, the state-of-the-art VM works for prominent processes involved in semiconductor manufacturing are surveyed, and the unique contributions of those VM predictive works are summarized and compared. Nevertheless, VM is still in its infancy stage, and there are significant hurdles that prevent wider VM implementation and acceptance in semiconductor industry. In this end, a critical analysis is provided to reveal the research gaps and discuss the challenges, such that further research can be simulated in the future. It is hoped that this work could illuminate researchers to understand the needs and future directions towards developing VM technology for semiconductor processes.
Virtual metrology of material removal rate using a one-dimensional convolutional neural network-based bidirectional long short-term memory network with attention
2023, Computers and Industrial Engineering
Material removal rate (MRR) is a key parameter for chemical mechanical polishing (CMP). It can affect the polished product quality, and has often been used as an indicator of CMP performance. Virtual Metrology (VM) is an technique for estimating the target value without physical measurement. It can reduce measurement time and save the cost of measurement tools. In this paper, we propose a VM framework, One-Dimensional Convolutional Neural Network-based Bidirectional Long Short-term Memory Network with Attention (1DCNN_BiLSTMwA), for monitoring the condition of components in a wafer chemical mechanical polishing system, to solve the complex data pre-process and feature extraction issue of traditional VM, and to achieve high accuracy of prediction. To evaluate the effectiveness of the proposed method, an empirical dataset from a semiconductor CMP process was used. According to the results of comparison, the proposed 1DCNN_BiLSTMwA has the lowest prediction error and outperforms 1DCNN, LSTM, BiLSTM models and the other machine learning techniques with statistical features and hand-designed features. A weighted, deep learning model was also used to improve performance. The experimental results demonstrate the practical viability of the proposed approach.
Development of a virtual metrology system for smart manufacturing: A case study of spandex fiber production
2023, Computers in Industry
Citation Excerpt :
To develop an advanced process control using VM, the working mechanism of the model needs to be comprehensible to multiple stakeholders, including data scientists, operators, regulators, and customers. However, few studies have adopted prediction models that incorporate algorithmic transparency, including multiple linear regression (Wu et al., 2017; Kim et al., 2017; Purwins et al., 2011), decision tree (Çiflikli and Kahya-Özyirmidokuz, 2010), partial least squares (Lynn et al., 2010;Purwins et al., 2011), fused least absolute shrinkage and selection operator (LASSO) (Park and Kim, 2016) approaches. This may be related to the fact that few VM models have been deployed in the real-world.
Virtual metrology (VM) is one of the most important enabling technologies in smart manufacturing. Although there is an abundance of literature on VM applications, the context of continuous production has received less attention. Fundamental challenges involved in the application of VM to a continuous process have been overlooked in comparison with intermittent (or batch) processes. Here, we described a real-world VM system for the manufacturing of spandex fiber, focusing on how practical challenges associated with a continuous process, including time synchronization, recirculation process, and autocorrelated features, can be addressed using data analytics. A model refresh strategy is discussed for the deployed VM system to ensure continuous usability and high-quality predictions. The virtual-physical connection established by the VM system creates a virtuous cycle in which constantly updated data render the system realistic, and valuable insights generated by the system can be applied to the physical production environment.
A data mining framework for financial prediction
2021, Expert Systems with Applications
Citation Excerpt :
Recently, data mining frameworks are proposed to solve domain-specific tasks. Kim et al. (2017) proposed a data mining framework of virtual metrology to provide high predictive performance and business-relevant information using data collected through sensors in the manufacturing domain. They identified optimal models by defining accuracy metrics for copper-clad laminate particularly designed for the manufacturing domain using several prediction models and dimensional reductions.
In the financial markets, because real-time transactions directly relate to profit, it is important to process and analyze data on a real-time basis. In practice, decisions influenced by experts’ experiences from fundamental and technical analysis occur frequently compared to decisions using prediction algorithms. A domain-specific data mining framework was proposed recently to reduce related cost. Therefore, this study proposes a novel data mining framework suitable for financial markets according to expert knowledge. The proposed framework predominantly considers the following three perspectives as the standards for the effectiveness of research: interpretability, proper prediction metrics, and reporting methods. We applied our framework to the real-world financial prediction problems, such as the 3–10 year treasury spread forecasts. Consequently, we achieved an 84% prediction performance on the spread prediction and used hierarchical information to provide additional insight. In addition, we obtained practical knowledge and synergies through extraction of critical variables that can be used as a quick and accurate data-driven decision making support tool by active agents in the real world.
Photosensitive hybrid polysilsesquioxanes for etching-free processing of flexible copper clad laminate
2021, Composites Science and Technology
Citation Excerpt :
And as the number signal I/O increases, smaller and smaller Cu interconnect linewidths are required for both reliability and speed [10]. A representative electronic material structure widely adopted for printed circuit board (PCB) applications are copper clad laminates (CCL), simply comprised of an insulating layer laminated with a Cu overlay [11,12]. A rigid CCL is composed of epoxy resins impregnated in glass fabrics [13], while flexible CCLs (FCCLs) utilize polyimide films as insulating substrate [14], and various additives further composited according to user needs.
A photosensitive ladder-structured polysilsesquioxane was synthesized for the development of flexible copper clad laminates through electroless plating of Cu. The developed ladder-structured polysilsesquioxane with controlled structure enabled easy introduction of protected thiol and allyl functional group, which under acidic conditions, the thiol group becomes deprotected. This deprotected thiol group enables UV-initiated thiol-ene curing of the exposed regions, while the unexposed regions containing free thiol groups act as seed initiators for Cu electroless plating, and the corresponding negative pattern fabrication process also possible. Through this facile method, we were able to develop flexible copper clad laminates of tunable Cu interconnect width size of high electrical conductivity according to photomask pattern and the insulating dielectric surrounding layer exhibiting low dielectric constant and low loss behavior without any corrosive Cu etching steps.
Multitask learning for virtual metrology in semiconductor manufacturing systems
2018, Computers and Industrial Engineering
Citation Excerpt :
As the dimensions of devices shrink, the time gap between measurements must also shrink (Chen et al., 2005). In response to these shortcomings, the concept of virtual metrology (VM; Fig. 2) has been developed (Chang, Kang, Hsu, Chang, & Chan, 2006; Kang et al., 2009; Kim et al., 2017). These metrology measurements relate to the process conditions.
Virtual metrology (VM) estimates the real metrology of wafers from process data collected from multiple chambers. In semiconductor manufacturing, independent models for each process chamber are limited because the number of sampled wafers measured at each chamber are too few to build a reliable model. One potential solution to this problem is to pool the data from all chambers to create a model capable of learning and serving as a global predictive model. However, even with chambers that perform the same operation, the condition of their semiconductor tools may vary because of various factors. This study uses, for the first time, various multitask methods to develop VM models. By learning multiple related tasks simultaneously, multitask methods effectively increase the number of observations included in the prediction model. In addition, by identifying the related task, the method can make a prediction using only similar tasks. This property of multitask learning can be useful to account for lack of information in a single chamber and for diversity among the chambers. The experimental results indicate that multitask models consistently outperformed independent and pooled models regardless of the size of the training set used. Among the multitask methods, a multitask tree-based ensemble model outperformed the others in every case. This implies that the problem of wafer quality prediction can be better addressed with a form of multitask learning.

View all citing articles on Scopus

View full text

Virtual metrology for copper-clad laminate manufacturing

Highlights

Abstract

Graphical abstract

Introduction

Section snippets

Related work

Virtual metrology

Data description and preprocessing

Conclusion

Acknowledgements

Materials Science in Semiconductor Processing

Applied Soft Computing

Decision Support Systems

Computers & Industrial Engineering

Computers & Industrial Engineering

Computers & Industrial Engineering

Expert Systems with Applications

Computers & Industrial Engineering

Computers & Industrial Engineering

Computers & Industrial Engineering

Control Engineering Practice

Computers & Industrial Engineering

Wafer-to-wafer virtual metrology applied to run-to-run control

Random forests

Machine Learning

Value-added metrology

IEEE Transactions on Semiconductor Manufacturing