Prahalad K. Rao

Online Real-Time Quality Monitoring in Additive Manufacturing Processes Using Heterogeneous Sensors

The objective of this work is to identify failure modes and detect the onset of process anomalies in additive manufacturing (AM) processes, specifically focusing on fused filament fabrication (FFF). We accomplish this objective using advanced Bayesian nonparametric analysis of in situ heterogeneous sensor data. Experiments are conducted on a desktop FFF machine instrumented with a heterogeneous sensor array including thermocouples, accelerometers, an infrared (IR) temperature sensor, and a real-time miniature video borescope. FFF process failures are detected online using the nonparametric Bayesian Dirichlet process (DP) mixture model and evidence theory (ET) based on the experimentally acquired sensor data. This sensor data-driven defect detection approach facilitates real-time identification and correction of FFF process drifts with an accuracy and precision approaching 85% (average F-score). In comparison, the F-score from existing approaches, such as probabilistic neural networks (PNN), naive Bayesian clustering, support vector machines (SVM), and quadratic discriminant analysis (QDA), was in the range of 55–75%.

Adaptive Neuro-Fuzzy Inference System Modeling of MRR and WIWNU in CMP Process With Sparse Experimental Data

Availability of only limited or sparse experimental data impedes the ability of current models of chemical mechanical planarization (CMP) to accurately capture and predict the underlying complex chemomechanical interactions. Modeling approaches that can effectively interpret such data are therefore necessary. In this paper, a new approach to predict the material removal rate (MRR) and within wafer nonuniformity (WIWNU) in CMP of silicon wafers using sparse-data sets is presented. The approach involves utilization of an adaptive neuro-fuzzy inference system (ANFIS) based on subtractive clustering (SC) of the input parameter space. Linear statistical models were used to assess the relative significance of process input parameters and their interactions. Substantial improvements in predicting CMP behaviors under sparse-data conditions can be achieved from fine-tuning membership functions of statistically less significant input parameters. The approach was also found to perform better than alternative neural network (NN) and neuro-fuzzy modeling methods for capturing the complex relationships that connect the machine and material parameters in CMP with MRR and WIWNU, as well as for predicting MRR and WIWNU in CMP.

Classification of atrial fibrillation episodes from sparse electrocardiogram data.

BACKGROUND Atrial fibrillation (AF) is the most common form of cardiac arrhythmia. This paper presents the application of the Classification and Regression Tree (CART) technique for detecting spontaneous termination or sustenance of AF with sparse data. METHOD Electrocardiogram (ECG) recordings were obtained from the PhysioNet (AF Termination Challenge Database 2004) Web site. Signal analysis, feature extraction, and classification were made to distinguish among 3 AF episodes, namely, Nonterminating (N), Soon (<1 minute) to be terminating (S), and Terminating immediately (<1 second) (T). RESULTS A continuous wavelet transform whose basis functions match the EKG patterns was found to yield compact representation (approximately 2 orders of magnitude). This facilitates the development of efficient algorithms for beat detection, QRST subtraction, and multiple ECG quantifier extraction (eg, QRS width, QT interval). A compact feature set was extracted through principal component analysis of these quantifiers. Accuracies exceeding 90% for AF episode classification were achieved. CONCLUSIONS A wavelet representation customized to the ECG signal pattern was found to yield 98% lower entropies compared with other representations that use standard library wavelets. The Classification and Regression Tree (CART) technique seems to distinguish the N vs T, and the S vs T classifications very accurately.

A graph-theoretic approach for quantification of surface morphology variation and its application to chemical mechanical planarization process

We present an algebraic graph-theoretic approach for quantification of surface morphology. Using this approach, heterogeneous, multi-scaled aspects of surfaces; e.g., semiconductor wafers, are tracked from optical micrographs as opposed to reticent profile mapping techniques. Therefore, this approach can facilitate in situ real-time assessment of surface quality. We report two complementary methods for realizing graph-theoretic representation and subsequent quantification of surface morphology variations from optical micrograph images. Experimental investigations with specular finished copper wafers (surface roughness (Sa) ∼ 6 nm) obtained using a semiconductor chemical mechanical planarization process suggest that the graph-based topological invariant Fiedler number (λ2) was able to quantify and track variations in surface morphology more effectively compared to other quantifiers reported in literature.

Assessment of Dimensional Integrity and Spatial Defect Localization in Additive Manufacturing Using Spectral Graph Theory

The ability of additive manufacturing (AM) processes to produce components with virtually any geometry presents a unique challenge in terms of quantifying the dimensional quality of the part. In this paper, a novel spectral graph theory (SGT) approach is proposed for resolving the following critical quality assurance concern in the AM: how to quantify the relative deviation in dimensional integrity of complex AM components. Here, the SGT approach is demonstrated for classifying the dimensional integrity of standardized test components. The SGT-based topological invariant Fiedler number (λ2) was calculated from 3D point cloud coordinate measurements and used to quantify the dimensional integrity of test components. The Fiedler number was found to differ significantly for parts originating from different AM processes (statistical significance p-value <1%). By comparison, prevalent dimensional integrity assessment techniques, such as traditional statistical quantifiers (e.g., mean and standard deviation) and examination of specific facets/landmarks failed to capture part-to-part variations, proved incapable of ranking the quality of test AM components in a consistent manner. In contrast, the SGT approach was able to consistently rank the quality of the AM components with a high degree of statistical confidence independent of sampling technique used. Consequently, from a practical standpoint, the SGT approach can be a powerful tool for assessing the dimensional integrity of the AM components, and thus encourage wider adoption of the AM capabilities.

Process-Machine Interaction (PMI) Modeling and Monitoring of Chemical Mechanical Planarization (CMP) Process Using Wireless Vibration Sensors

We present a deterministic process-machine interaction (PMI) model that can associate different complex time-frequency patterns, including nonlinear dynamic behaviors that manifest in vibration signals measured during a chemical mechanical planarization (CMP) process for polishing blanket copper wafer surfaces to near-optical finish (Ra ~ 5 nm) to specific process mechanisms. The model captures the effects of the nonuniform structural properties of the polishing pad, pad asperities, and machine kinematics on CMP dynamics using a deterministic 2 ° of freedom nonlinear differential equation. The model was validated using a Buehler (Automet 250) bench top CMP machine instrumented with a wireless (XBee IEEE 802.15.4 RF module) multi-sensor unit that includes a MEMS 3-axis accelerometer (Analog Devices ADXL 335). Extensive experiments suggest that the deterministic PMI model can capture such significant signal patterns as aperiodicity, broadband frequency spectra, and other prominent manifestations of process nonlinearity. Remarkably, the deterministic PMI model was able to explain not just the physical sources of various time-frequency patterns observed in the measured vibration signals, but also, their variations with process conditions. The features extracted from experimental vibration data, such as power spectral density over the 115-120 Hz band, and nonlinear recurrence measures were statistically significant estimators (R2 ~ 75%) of process parameter settings. The model together with sparse experimental data was able to estimate process drifts resulting from pad wear with high fidelity (R2 ~ 85%). The signal features identified using the PMI model can lead to effective real-time in-situ monitoring of wear and anomalies in the CMP process.

An online sparse estimation-based classification approach for real-time monitoring in advanced manufacturing processes from heterogeneous sensor data

ABSTRACT The objective of this work is to realize real-time monitoring of process conditions in advanced manufacturing using multiple heterogeneous sensor signals. To achieve this objective we propose an approach invoking the concept of sparse estimation called online sparse estimation-based classification (OSEC). The novelty of the OSEC approach is in representing data from sensor signals as an underdetermined linear system of equations and subsequently solving the underdetermined linear system using a newly developed greedy Bayesian estimation method. We apply the OSEC approach to two advanced manufacturing scenarios, namely, a fused filament fabrication additive manufacturing process and an ultraprecision semiconductor chemical–mechanical planarization process. Using the proposed OSEC approach, process drifts are detected and classified with higher accuracy compared with popular machine learning techniques. Process drifts were detected and classified with a fidelity approaching 90% (F-score) using OSEC. In comparison, conventional signal analysis techniques—e.g., neural networks, support vector machines, quadratic discriminant analysis, naïve Bayes—were evaluated with F-score in the range of 40% to 70%.

Classifying the Dimensional Variation in Additive Manufactured Parts From Laser-Scanned Three-Dimensional Point Cloud Data Using Machine Learning Approaches

The objective of this work is to develop and apply a spectral graph theoretic approach for differentiating between (classifying) additive manufactured (AM) parts contingent on the severity of their dimensional variation from laser-scanned coordinate measurements (3D point cloud). The novelty of the approach is in invoking spectral graph Laplacian eigenvalues as an extracted feature from the laser-scanned 3D point cloud data in conjunction with various machine learning techniques. The outcome is a new method that classifies the dimensional variation of an AM part by sampling less than 5% of the 2 million 3D point cloud data acquired (per part). This is a practically important result, because it reduces the measurement burden for postprocess quality assurance in AM— parts can be laser-scanned and their dimensional variation quickly assessed on the shop floor. To realize the research objective, the procedure is as follows. Test parts are made using the fused filament fabrication (FFF) polymer AM process. The FFF process conditions are varied per a phased design of experiments plan to produce parts with distinctive dimensional variations. Subsequently, each test part is laser scanned and 3D point cloud data are acquired. To classify the dimensional variation among parts, Laplacian eigenvalues are extracted from the 3D point cloud data and used as features within different machine learning approaches. Six machine learning approaches are juxtaposed: sparse representation, k-nearest neighbors, neural network, na€ıve Bayes, support vector machine, and decision tree. Of these, the sparse representation technique provides the highest classification accuracy (F-score> 97%). [DOI: 10.1115/1.4036641]

Sensor-Based Online Monitoring and Computational Fluid Dynamics (CFD) Modeling of Aerosol Jet Printing (AJP) Process

The objectives of this paper in the context of aerosol jet printing (AJP)—an additive manufacturing (AM) process—are to: (1) realize in situ online monitoring of print quality in terms of line/electronic trace morphology; and (2) explain the causal aerodynamic interactions that govern line morphology based on a two-dimensional computational fluid dynamics (2D-CFD) model. To realize these objectives, an Optomec AJ-300 aerosol jet printer was instrumented with a charge coupled device (CCD) camera mounted coaxial to the nozzle (perpendicular to the platen). Experiments were conducted by varying two process parameters, namely, sheath gas flow rate (ShGFR) and carrier gas flow rate (CGFR). The morphology of the deposited lines was captured from the online CCD images. Subsequently, using a novel digital image processing method proposed in this study, six line morphology attributes were quantified. The quantified line morphology attributes are: (1) line width, (2) line density, (3) line edge quality/smoothness, (4) overspray (OS), (5) line discontinuity, and (6) internal connectivity. The experimentally observed line morphology trends as a function of ShGFR and CGFR were verified with computational fluid dynamics (CFD) simulations. The image-based line morphology quantifiers proposed in this work can be used for online detection of incipient process drifts, while the CFD model is valuable to ascertain the appropriate corrective action to bring the process back in control in case of a drift. [DOI: 10.1115/1.4034591]

Heterogeneous Sensor Data Fusion Approach for Real-time Monitoring in Ultraprecision Machining (UPM) Process Using Non-Parametric Bayesian Clustering and Evidence Theory

The aim of this paper is to detect the incipient anomalies in a ultraprecision machining (UPM) process by integrating multiple in situ sensor signals. To realize this aim we forward a Bayesian non-parametric Dirichlet Process (DP) decision-making approach for real-time monitoring of UPM process using the data gathered from multiple, heterogeneous sensors. The sensor signals are acquired under different experimental conditions from a UPM setup instrumented with a heterogeneous sensing array consisting of miniature tri-axis force, tri-axis vibration, and acoustic emission (AE) sensors mounted in close proximity to the cutting tool. We track the prominent nonlinear and non-Gaussian signal patterns evident in the experimentally acquired sensor data using an adaptive non-parametric DP modeling technique. A cohesive decision concerning the UPM process condition is made by developing a new supervised learning method, which integrates the DP-model state estimates with an evidence theoretic sensor data fusion method. Using this combined DP-evidence theoretic approach, UPM process drifts and anomalies, such as sudden changes in the depth of cut, feed rate, and spindle speed that deleteriously affect surface finish, and hence cause high yield losses, are detected and classified with over 90% accuracy (with % standard deviation). We compared these results with popular classification techniques, e.g., naïve Bayes, self-organizing map, and support vector machine; these conventional techniques had classification accuracy in the range of 83%-88%. Consequently, this research makes the following practically relevant contributions: 1) real-time identification of the incipient UPM process anomalies from multiple sensors and 2) prescribing the optimal subset of sensors signals contingent to particular process anomalies. This paper addresses the critical problem of real-time monitoring in an ultraprecision machining (UPM) process, also called diamond turning (DT). The main contribution of this work is the development of a novel analytical approach that combines signals from multiple in situ sensors for detecting abnormal process drifts in UPM. We show that this approach, which invokes a powerful mathematical technique called Dirichlet Process (DP), is capable of classifying UPM process drifts much more accurately than conventional classification methods reported in the literature (e.g., neural networks, and support vector machines). As a practical consequence of this research, an UPM operator can infer, based on sensor data, the occurrence of process faults, as well as isolate the type of fault. This is practically valuable because appropriate and opportune corrective action can be taken to avoid the heavy scrap and rework rates endemic to UPM.