Chemical process plant safety, production specifications, environmental regulations, operational constraints, and plant economics are some of the main reasons driving an upward interest in research and development of more robust methods for process monitoring and control. Principal component analysis (PCA) has long been used in fault detection by extracting relevant information from multivariate chemical data. The recent success of wavelets and multi-scale methods in chemical process monitoring and control has catalyzed an interest in the investigation of wavelets based methods for fault detection. In the present work, multi-scale principal component analysis (MSPCA) is used for fault detection and diagnosis. MSPCA simultaneously extracts both, cross correlation across the sensors (PCA approach) and auto-correlation within a sensor (wavelet approach). Using wavelets, the individual sensor signals are decomposed into approximations and details at different scales. Contributions from each scale are collected in separate matrices, and a PCA model is then constructed to extract correlation at each scale. The multi-scale nature of MSPCA formulation makes it suitable to work with process data that are typically non-stationary and represent the cumulative effect of many underlying process phenomena, each operating at a different scale. The proposed MSPCA approach is able to outperform the conventional PCA based approach in detecting and identifying real process faults in an industrial process, and yields minimum false alarms. Additionally, the advantage of MSPCA, over the traditional PCA approach for sensor validation, is also demonstrated on an industrial boiler data set.

This article proposes a new approach for etch endpoint detection of small open area wafers. The traditional endpoint detection technique uses a few manually selected wavelengths, which are adequate for large open areas. As the integrated circuit devices continue to shrink in geometry and increase in device density, detecting the endpoint for small open areas presents a serious challenge to process engineers. In this work, a high-resolution optical emission spectroscopy (OES) system is used to provide the necessary sensitivity for detecting subtle endpoint signals. Principal component analysis is used to analyze the OES data and extract key components that capture the endpoint signal. Data analysis from many wafers shows that the endpoint pattern in the principal components is repeatable. Two methods are used to select wavelengths so as to improve reliability and reduce susceptibility to noise. The first method is to remove those spectrum windows that contain no endpoint information. The second method is to use a “sphere” criterion to select the most relevant wavelengths. The final endpoint algorithm using a much-reduced number of wavelengths shows more distinguishable and reliable endpoint features.

Process monitoring and fault diagnosis are crucial for efficient and optimal operation of a chemical plant. This paper proposes a reconstruction-based fault identification approach using a combined index for multidimensional fault reconstruction and identification. Fault detection is conducted using a new index that combines the squared prediction error (SPE) and T-2. Necessary and sufficient conditions for fault detectability are derived. The combined index is used to reconstruct the fault along a given fault direction. Faults are identified by assuming that each fault in a candidate fault set is the true fault and comparing the reconstructed indices with the control limits. Fault reconstructability and identifiability on the basis of the combined index are discussed. A new method to extract fault directions from historical fault data is proposed. The dimension of the fault is determined on the basis of the fault detection indices after fault reconstruction. Several simulation examples and one practical case are presented. The method proposed here is compared with two existing methods in the literature for the identification single-sensor and multiple-sensor faults. We analyze the reasons that the other two methods lead to erroneous identification results. Finally, the proposed approach is applied to a rapid thermal annealing process for fault diagnosis. Fault subspaces of several typical process faults are extracted from the data and then used to identify new faults.

The objective of this paper is to investigate the suitability of using optical emission spectroscopy (OES) for the fault detection and classification of plasma etchers. The OES sensor system used in this study can collect spectra at up to 512 different wavelengths, Multiple scans of the spectra are taken from a wafer, and the spectra data are available for multiple wafers, As a result, the amount of the OES data is typically large, This poses a difficulty in extracting relevant information for fault detection and classification, In this paper, we propose the use of multiway principal component analysis (PCA) to analyze the sensitivity of the multiple scans within a wafer with respect to typical faults such as etch stop, which is a fault that occurs when the polymer deposition rate is larger than the etch rate, Several PCA-based schemes are tested for the purpose of fault detection and wavelength selection, A sphere criterion is proposed for wavelength selection and compared with an existing method in the literature. To construct the final monitoring model, the OES data of selected wavelengths are properly scaled to calculate fault detection indices, Reduction in the number of wavelengths implies reduced cost for implementing the fault detection system. All experiments are conducted on an Applied Materials 5300 oxide etcher at Advanced Micro Devices (AMD) in Austin, TX.

While principal component analysis (PCA) has found wide application in process monitoring, slow and normal process changes often occur in real processes, which lead to false alarms for a fixed-model monitoring approach. In this paper, we propose two recursive PCA algorithms for adaptive process monitoring. The paper starts with an efficient approach to updating the correlation matrix recursively. The algorithms, using rank-one modification and Lanczos tridiagonalization, are then proposed and their computational complexity is compared. The number of principal components and the confidence limits for process monitoring are also determined recursively. A complete adaptive monitoring algorithm that addresses the issues of missing values and outlines is presented. Finally, the proposed algorithms are applied to a rapid thermal annealing process in semiconductor processing for adaptive monitoring.