Semiconductor Manufacturing Data Research Articles

Quantum mechanics-based missing value estimation framework for industrial data

As the importance of data-based predictive maintenance frameworks is rising, one of the emerging issues is the missing values in industrial data. Data with various missing values due to measurement sensor faults and power change issues can not only cause errors in future analyses but also lead to data shortages. Although various missing value estimation methods have been proposed to address this issue, correction methods that consider the fine uncertainty and noise of the missing values themselves have been relatively unexplored. Therefore, this study modeled missing values estimated using Gaussian progress regression by considering the fine movements and noises of data using a quantum mechanics-based stochastic differential equation and corrected them using Ito’s lemma. Estimating the missing data in this manner more closely simulates the attributes of data compared to existing estimation methods, enabling more accurate analysis. To demonstrate the excellence of the proposed framework, missing values were estimated using vehicle operation data and semiconductor manufacturing data with multiple missing values, and the estimation results were compared with those of existing algorithms.

Bayesian Nonparametric Classification for Incomplete Data With a High Missing Rate: an Application to Semiconductor Manufacturing Data

During the semiconductor manufacturing process, predicting the yield of the semiconductor is an important problem. Early detection of defective product production in the manufacturing process can save huge production cost. The data generated from the semiconductor manufacturing process have characteristics of non-normal distributions, random missing patterns and high missing rate, which complicate the prediction of the yield. We propose the Dirichlet Process -Naive Bayes model (DPNB) that can simultaneously impute missing values and address classification problems. Since the DPNB is based on the infinite Gaussian mixture model, it can estimate complex data distributions and make predictions for missing datasets with some missing patterns due to nice properties of the Gaussian distribution. The DPNB also performs well for high missing rates since it uses all information of observed components. Experiments on various real datasets including semiconductor manufacturing data show that the DPNB has better performance than state-of-the-art methods in terms of predicting missing values and target variables as percentage of missing values increases.

반도체 계측 데이터 기반 군집화를 활용한 개선된 품질 예측 방법론

Various methods have been applied to guarantee and improve the quality of products in the semiconductor manufacturing. However, defects are becoming more various and difficult to control with product diversification and technology advancement. To ensure and improve the quality with productivity, this study predicted the final quality with actual semiconductor manufacturing data generated in each process of various characteristics. To improve the performance with practicality, failure occurrence environment and data characteristics should be considered. As the technology complexity increases, the defect frequently occurs with a same phenomenon but different root cause. Therefore, we proposed the system that divides defect types by characteristics and predict quality with unsupervised learning such as k-means and SOM (Self-Organized Map). The proposed method could provide an individual clue to improvements by clustering characteristics for defects. In addition, it showed verified applicability by improving performance about 4.4%p in AUC (Area Under the ROC Curve) and 6.8%p in partial AUC.

An Experimental Evaluation of Fault Diagnosis from Imbalanced and Incomplete Data for Smart Semiconductor Manufacturing

The SECOM dataset contains information about a semiconductor production line, entailing the products that failed the in-house test line and their attributes. This dataset, similar to most semiconductor manufacturing data, contains missing values, imbalanced classes, and noisy features. In this work, the challenges of this dataset are met and many different approaches for classification are evaluated to perform fault diagnosis. We present an experimental evaluation that examines 288 combinations of different approaches involving data pruning, data imputation, feature selection, and classification methods, to find the suitable approaches for this task. Furthermore, a novel data imputation approach, namely “In-painting KNN-Imputation” is introduced and is shown to outperform the common data imputation technique. The results show the capability of each classifier, feature selection method, data generation method, and data imputation technique, with a full analysis of their respective parameter optimizations.

Bayesian inference for mining semiconductor manufacturing big data for yield enhancement and smart production to empower industry 4.0

Big data analytics have been employed to extract useful information and derive effective manufacturing intelligence for yield management in semiconductor manufacturing that is one of the most complex manufacturing processes due to tightly constrained production processes, reentrant process flows, sophisticated equipment, volatile demands, and complicated product mix. Indeed, the increasing adoption of multimode sensors, intelligent equipment, and robotics have enabled the Internet of Things (IOT) and big data analytics for semiconductor manufacturing. Although the processing tool, chamber set, and recipe are selected according to product design and previous experiences, domain knowledge has become less efficient for defect diagnosis and fault detection. To fill the gaps, this study aims to develop a framework based on Bayesian inference and Gibbs sampling to investigate the intricate semiconductor manufacturing data for fault detection to empower intelligent manufacturing. In addition, Cohen’s kappa coefficient was used to eliminate the influence of extraneous variables. The proposed approach was validated through an empirical study and simulation. The results have shown the practical viability of the proposed approach.

Efficient Feature Selection-Based on Random Forward Search for Virtual Metrology Modeling

Virtual metrology (VM) has been successfully applied to semiconductor manufacturing as an efficient way of achieving wafer-to-wafer quality control. VM involves the estimation of metrology variables of wafer inspection using a prediction model trained with process parameters and measurements prior to the actual implementation of metrology. VM modeling should incorporate a number of process parameters and measurements collected from each piece of process equipment, which results in a greater number of input variables. Therefore, it is necessary to resolve the problem of high dimensionality through feature selection. A suitable feature selection method for VM modeling should effectively address the high dimensionality by lowering the computational cost, while also achieving high prediction accuracy as an essential requirement for the practical deployment of VM. In this paper, a feature selection method based on random forward search is proposed for efficient VM modeling. This method selects relevant variables sequentially from disjoint random subsets of candidate variables by incorporating randomness. Our preliminary experimental results obtained with real-world semiconductor manufacturing data demonstrate that the proposed feature selection method achieves comparable prediction accuracy yet has the advantages of being computationally more efficient, thus merits further investigation.

Machine learning-based novelty detection for faulty wafer detection in semiconductor manufacturing

Since semiconductor manufacturing consists of hundreds of processes, a faulty wafer detection system, which allows for earlier detection of faulty wafers, is required. statistical process control (SPC) and virtual metrology (VM) have been used to detect faulty wafers. However, there are some limitations in that SPC requires linear, unimodal and single variable data and VM underestimates the deviations of predictors. In this paper, seven different machine learning-based novelty detection methods were employed to detect faulty wafers. The models were trained with Fault Detection and Classification (FDC) data to detect wafers having faulty metrology values. The real world semiconductor manufacturing data collected from a semiconductor fab were tested. Since the real world data have more than 150 input variables, we employed three different dimensionality reduction methods. The experimental results showed a high True Positive Rate (TPR). These results are promising enough to warrant further study.

Data mining for yield enhancement in semiconductor manufacturing and an empirical study

During wafer fabrication, process data, equipment data, and lot history will be automatically or semi-automatically recorded and accumulated in database for monitoring the process, diagnosing faults, and managing manufacturing. However, in high-tech industry such as semiconductor manufacturing, many factors that are interrelated affect the yield of fabricated wafers. Engineers who rely on personal domain knowledge cannot find possible root causes of defects rapidly and effectively. This study aims to develop a framework for data mining and knowledge discovery from database that consists of a Kruskal–Wallis test, K-means clustering, and the variance reduction splitting criterion to investigate the huge amount of semiconductor manufacturing data and infer possible causes of faults and manufacturing process variations. The extracted information and knowledge is helpful to engineers as a basis for trouble shooting and defect diagnosis. We validated this approach with an empirical study in a semiconductor foundry company in Taiwan and the results demonstrated the practical viability of this approach.

Mining semiconductor manufacturing data for productivity improvement — an integrated relational database approach

The semiconductor manufacturing process is a very complex process due to the large number of processes, the diverse equipment set, and the complexities associated with its non-linear process flows. Today, management seeks effective real time reporting on different key parameters such as work-in-process (WIP), cycle time, scrap, and tool up-time in an effort to monitor factory performance so that the business goals are met. The complexity of wafer fabrication (fab) processing, however, has necessitated the need for multiple multi-million dollar Computer Integrated Manufacturing (CIM) systems to be implemented in order to collect real time data. CIM systems deployed in semiconductor manufacturing settings capture real time data about every detail of the factory such as equipment throughputs, WIP, and product and equipment history. Most of the data collected, however, tends to be archived rather than used due to the difficulty of data extraction and presentation. The challenge, therefore, is to transform the massive data in the different CIM databases into meaningful information that is easy to access as well as easy to interpret and manipulate. This paper presents an integrated relational database approach for modeling and collecting semiconductor manufacturing data from multiple database systems and transforming the data into useful reports. A summary of the key reports generated is presented as well. These reports help to monitor factory performance by tracking different key metrics. These reports were implemented in one of Motorola’s wafer fabs and have contributed significantly in improving factory performance.

Knowledge acquisition from batch semiconductor manufacturing data

Semiconductor manufacturing data consist of the processes and the machines involved in the production of batches of semiconductor circuit wafers. Wafer quality depends on the manufacturing line status and it is measured at the end of the line. We have developed a knowledge discovery system that is intended to help the yield analysis expert by learning the tentative causes of low quality wafers from an exhaustive amount of manufacturing data. The yield analysis expert, by using the knowledge discovered, will decide on which corrective actions to perform on the manufacturing process. This paper discusses the transformations carried out within the data from raw data to discovered knowledge, and also the two main tasks performed by the system. The features of the inductive algorithm performing those tasks are also described. Yield analysis experts at Lucent Technologies, Bell Labs Innovations in Spain are currently using this knowledge discovery application.

Knowledge acquisition from batch semiconductor manufacturing data

Semiconductor manufacturing data consist of the processes and the machines involved in the production of batches of semiconductor circuit wafers. Wafer quality depends on the manufacturing line status and it is measured at the end of the line. We have developed a knowledge discovery system that is intended to help the yield analysis expert by learning the tentative causes of low quality wafers from an exhaustive amount of manufacturing data. The yield analysis expert, by using the knowledge discovered, will decide on which corrective actions to perform on the manufacturing process. This paper discusses the transformations carried out within the data from raw data to discovered knowledge, and also the two main tasks performed by the system. The features of the inductive algorithm performing those tasks are also described. Yield analysis experts at Lucent Technologies, Bell Labs Innovations in Spain are currently using this knowledge discovery application.