Feature Selection for Anomaly Detection Using Optical Emission Spectroscopy

To maintain the pace of development set by Moore's law, production processes in semiconductor manufacturing are becoming more and more complex. The development of efficient and interpretable anomaly detection systems is fundamental to keeping production costs low. As the dimension of process monitoring data can become extremely high anomaly detection systems are impacted by the curse of dimensionality, hence dimensionality reduction plays an important role. Classical dimensionality reduction approaches, such as Principal Component Analysis, generally involve transformations that seek to maximize the explained variance. In datasets with several clusters of correlated variables the contributions of isolated variables to explained variance may be insignificant, with the result that they may not be included in the reduced data representation. It is then not possible to detect an anomaly if it is only reflected in such isolated variables. In this paper we present a new dimensionality reduction technique that takes account of such isolated variables and demonstrate how it can be used to build an interpretable and robust anomaly detection system for Optical Emission Spectroscopy data.

FPGA-based architecture for hyperspectral endmember extraction

Hyperspectral instruments have been incorporated in satellite missions, providing data of high spectral resolution of the Earth. This data can be used in remote sensing applications, such as, target detection, hazard prevention, and monitoring oil spills, among others. In most of these applications, one of the requirements of paramount importance is the ability to give real-time or near real-time response. Recently, onboard processing systems have emerged, in order to overcome the huge amount of data to transfer from the satellite to the ground station, and thus, avoiding delays between hyperspectral image acquisition and its interpretation. For this purpose, compact reconfigurable hardware modules, such as field programmable gate arrays (FPGAs) are widely used. This paper proposes a parallel FPGA-based architecture for endmember’s signature extraction. This method based on the Vertex Component Analysis (VCA) has several advantages, namely it is unsupervised, fully automatic, and it works without dimensionality reduction (DR) pre-processing step. The architecture has been designed for a low cost Xilinx Zynq board with a Zynq-7020 SoC FPGA based on the Artix-7 FPGA programmable logic and tested using real hyperspectral data sets collected by the NASA’s Airborne Visible Infra-Red Imaging Spectrometer (AVIRIS) over the Cuprite mining district in Nevada. Experimental results indicate that the proposed implementation can achieve real-time processing, while maintaining the methods accuracy, which indicate the potential of the proposed platform to implement high-performance, low cost embedded systems, opening new perspectives for onboard hyperspectral image processing.

Signal subspace identification in hyperspectral imagery

Terrestrial remote sensing imagery involves the acquisition of information from the Earth's surface without physical contact with the area under study. Among the remote sensing modalities, hyperspectral imaging has recently emerged as a powerful passive technology. This technology has been widely used in the fields of urban and regional planning, water resource management, environmental monitoring, food safety, counterfeit drugs detection, oil spill and other types of chemical contamination detection, biological hazards prevention, and target detection for military and security purposes [2-9]. Hyperspectral sensors sample the reflected solar radiation from the Earth surface in the portion of the spectrum extending from the visible region through the near-infrared and mid-infrared (wavelengths between 0.3 and 2.5 µm) in hundreds of narrow (of the order of 10 nm) contiguous bands [10]. This high spectral resolution can be used for object detection and for discriminating between different objects based on their spectral xharacteristics [6]. However, this huge spectral resolution yields large amounts of data to be processed. For example, the Airbone Visible/Infrared Imaging Spectrometer (AVIRIS) [11] collects a 512 (along track) X 614 (across track) X 224 (bands) X 12 (bits) data cube in 5 s, corresponding to about 140 MBs. Similar data collection ratios are achieved by other spectrometers [12]. Such huge data volumes put stringent requirements on communications, storage, and processing. The problem of signal sbspace identification of hyperspectral data represents a crucial first step in many hypersctral processing algorithms such as target detection, change detection, classification, and unmixing. The identification of this subspace enables a correct dimensionality reduction (DR) yelding gains in data storage and retrieval and in computational time and complexity. Additionally, DR may also improve algorithms performance since it reduce data dimensionality without losses in the useful signal components. The computation of statistical estimates is a relevant example of the advantages of DR, since the number of samples required to obtain accurate estimates increases drastically with the dimmensionality of the data (Hughes phnomenon) [13].

Implementazione e benchmarking dell'algoritmo QDANet PRO per l'analisi di big data genomici

Dato il recente avvento delle tecnologie NGS, in grado di sequenziare interi genomi umani in tempi e costi ridotti, la capacità di estrarre informazioni dai dati ha un ruolo fondamentale per lo sviluppo della ricerca. Attualmente i problemi computazionali connessi a tali analisi rientrano nel topic dei Big Data, con databases contenenti svariati tipi di dati sperimentali di dimensione sempre più ampia. Questo lavoro di tesi si occupa dell'implementazione e del benchmarking dell'algoritmo QDANet PRO, sviluppato dal gruppo di Biofisica dell'Università di Bologna: il metodo consente l'elaborazione di dati ad alta dimensionalità per l'estrazione di una Signature a bassa dimensionalità di features con un'elevata performance di classificazione, mediante una pipeline d'analisi che comprende algoritmi di dimensionality reduction. Il metodo è generalizzabile anche all'analisi di dati non biologici, ma caratterizzati comunque da un elevato volume e complessità, fattori tipici dei Big Data. L'algoritmo QDANet PRO, valutando la performance di tutte le possibili coppie di features, ne stima il potere discriminante utilizzando un Naive Bayes Quadratic Classifier per poi determinarne il ranking. Una volta selezionata una soglia di performance, viene costruito un network delle features, da cui vengono determinate le componenti connesse. Ogni sottografo viene analizzato separatamente e ridotto mediante metodi basati sulla teoria dei networks fino all'estrapolazione della Signature finale. Il metodo, già precedentemente testato su alcuni datasets disponibili al gruppo di ricerca con riscontri positivi, è stato messo a confronto con i risultati ottenuti su databases omici disponibili in letteratura, i quali costituiscono un riferimento nel settore, e con algoritmi già esistenti che svolgono simili compiti. Per la riduzione dei tempi computazionali l'algoritmo è stato implementato in linguaggio C++ su HPC, con la parallelizzazione mediante librerie OpenMP delle parti più critiche.

Dimensionality hyper-reduction and machine learning for dynamical systems with varying parameters

Thesis (Ph.D.)--University of Washington, 2016-08

VAM Applied to Dimensional Reduction of Nonlinear Multifunctional Film-fabric Laminates

This work aims at asymptotically accurate dimensional reduction of non-linear multi-functional film-fabric laminates having specific application in design of envelopes for High Altitude Airships (HAA). The film-fabric laminate for airship envelope consists of a woven fabric core coated with thin films on each face. These films provide UV protection and Helium leakage prevention, while the core provides required structural strength. This problem is both geometrically and materially non-linear. To incorporate the geometric non-linearity, generalized warping functions are used and finite deformations are allowed. The material non-linearity is handled by using hyper-elastic material models for each layer. The development begins with three-dimensional (3-D) nonlinear elasticity and mathematically splits the analysis into a one-dimensional through-the-thickness analysis and a two-dimensional (2-D) plate analysis. The through-the-thickness analysis provides the 2-D constitutive law which is then given as an input to the 2-D reference surface analysis. The dimensional reduction is carried out using Variational Asymptotic Method (VAM) for moderate strains and very small thickness-to-wavelength ratio. It features the identification and utilization of additional small parameters such as ratio of thicknesses and stiffness coefficients of core and films. Closed form analytical expressions for warping functions and 2-D constitutive law of the film-fabric laminate are obtained.

Cyclic Pursuit: Symmetry, Reduction and Nonlinear Dynamics

In this dissertation, we explore the use of pursuit interactions as a building block for collective behavior, primarily in the context of constant bearing (CB) cyclic pursuit. Pursuit phenomena are observed throughout the natural environment and also play an important role in technological contexts, such as missile-aircraft encounters and interactions between unmanned vehicles. While pursuit is typically regarded as adversarial, we demonstrate that pursuit interactions within a cyclic pursuit framework give rise to seemingly coordinated group maneuvers. We model a system of agents (e.g. birds, vehicles) as particles tracing out curves in the plane, and illustrate reduction to the shape space of relative positions and velocities. Introducing the CB pursuit strategy and associated pursuit law, we consider the case for which agent i pursues agent i+1 (modulo n) with the CB pursuit law. After deriving closed-loop cyclic pursuit dynamics, we demonstrate asymptotic convergence to an invariant submanifold (corresponding to each agent attaining the CB pursuit strategy), and proceed by analysis of the reduced dynamics restricted to the submanifold. For the general setting, we derive existence conditions for relative equilibria (circling and rectilinear) as well as for system trajectories which preserve the shape of the collective (up to similarity), which we refer to as pure shape equilibria. For two illustrative low-dimensional cases, we provide a more comprehensive analysis, deriving explicit trajectory solutions for the two-particle "mutual pursuit" case, and detailing the stability properties of three-particle relative equilibria and pure shape equilibria. For the three-particle case, we show that a particular choice of CB pursuit parameters gives rise to remarkable almost-periodic trajectories in the physical space. We also extend our study to consider CB pursuit in three dimensions, deriving a feedback law for executing the CB pursuit strategy, and providing a detailed analysis of the two-particle mutual pursuit case. We complete the work by considering evasive strategies to counter the motion camouflage (MC) pursuit law. After demonstrating that a stochastically steering evader is unable to thwart the MC pursuit strategy, we propose a (deterministic) feedback law for the evader and demonstrate the existence of circling equilibria for the closed-loop pursuer-evader dynamics.

Nonlinear MHD stability of aluminium reduction cells

An industrial electrolysis cell used to produce primary aluminium is sensitive to waves at the interface of liquid aluminium and electrolyte. The interface waves are similar to stratified sea layers [1], but the penetrating electric current and the associated magnetic field are intricately involved in the oscillation process, and the observed wave frequencies are shifted from the purely hydrodynamic ones [2]. The interface stability problem is of great practical importance because the electrolytic aluminium production is a major electrical energy consumer, and it is related to environmental pollution rate. The stability analysis was started in [3] and a short summary of the main developments is given in [2]. Important aspects of the multiple mode interaction have been introduced in [4], and a widely used linear friction law first applied in [5]. In [6] a systematic perturbation expansion is developed for the fluid dynamics and electric current problems permitting reduction of the three-dimensional problem to a two dimensional one. The procedure is more generally known as “shallow water approximation” which can be extended for the case of weakly non-linear and dispersive waves. The Boussinesq formulation permits to generalise the problem for non-unidirectionally propagating waves accounting for side walls and for a two fluid layer interface [1]. Attempts to extend the electrolytic cell wave modelling to the weakly nonlinear case have started in [7] where the basic equations are derived, including the nonlinearity and linear dispersion terms. An alternative approach for the nonlinear numerical simulation for an electrolysis cell wave evolution is attempted in [8 and references there], yet, omitting the dispersion terms and without a proper account for the dissipation, the model can predict unstable waves growth only. The present paper contains a generalisation of the previous non linear wave equations [7] by accounting for the turbulent horizontal circulation flows in the two fluid layers. The inclusion of the turbulence model is essential in order to explain the small amplitude self-sustained oscillations of the liquid metal surface observed in real cells, known as “MHD noise”. The fluid dynamic model is coupled to the extended electromagnetic simulation including not only the fluid layers, but the whole bus bar circuit and the ferromagnetic effects [9].

Reduction of nonlinear intrachannel effects by channel asymmetry in transmission lines with strong bit overlapping

We have examined the statistics of simulated bit-error rates in optical transmission systems with strong patterning effects and have found strong correlation between the probability of marks in a pseudorandom pattern and the error-free transmission distance. We discuss how a reduced density of marks can be achieved by preencoding optical data.

Fiber nonlinearity-induced penalty reduction in CO-OFDM by ANN-based nonlinear equalization

We experimentally demonstrate ∼2 dB quality (Q)-factor enhancement in terms of fiber nonlinearity compensation of 40 Gb/s 16 quadrature amplitude modulation coherent optical orthogonal frequency-division multiplexing at 2000 km, using a nonlinear equalizer (NLE) based on artificial neural networks (ANN). Nonlinearity alleviation depends on escalation of the ANN training overhead and the signal bit rate, reporting ∼4 dB Q-factor enhancement at 70 Gb/s, whereas a reduction of the number of ANN neurons annihilates the NLE performance. An enhanced performance by up to ∼2 dB in Q-factor compared to the inverse Volterra-series transfer function NLE leads to a breakthrough in the efficiency of ANN.

Model reduction for nonlinear aerodynamics and aeroelasticity using a Discrete Empirical Interpolation Method

A novel surrogate model is proposed in lieu of Computational Fluid Dynamics (CFD) solvers, for fast nonlinear aerodynamic and aeroelastic modeling. A nonlinear function is identified on selected interpolation points by
a discrete empirical interpolation method (DEIM). The flow field is then reconstructed using a least square approximation of the flow modes extracted
by proper orthogonal decomposition (POD). The aeroelastic reduce order
model (ROM) is completed by introducing a nonlinear mapping function
between displacements and the DEIM points. The proposed model is investigated to predict the aerodynamic forces due to forced motions using
a N ACA 0012 airfoil undergoing a prescribed pitching oscillation. To investigate aeroelastic problems at transonic conditions, a pitch/plunge airfoil
and a cropped delta wing aeroelastic models are built using linear structural models. The presence of shock-waves triggers the appearance of limit
cycle oscillations (LCO), which the model is able to predict. For all cases
tested, the new ROM shows the ability to replicate the nonlinear aerodynamic forces, structural displacements and reconstruct the complete flow
field with sufficient accuracy at a fraction of the cost of full order CFD
model.

Model reduction for nonlinear aeroelasticity using the Discrete Empirical Interpolation Method

A novel surrogate model is proposed in lieu of computational fluid dynamic (CFD) code for fast nonlinear aerodynamic modeling. First, a nonlinear function is identified on selected interpolation points defined by discrete empirical interpolation method (DEIM). The flow field is then reconstructed by a least square approximation of flow modes extracted by proper orthogonal decomposition (POD). The proposed model is applied in the prediction of limit cycle oscillation for a plunge/pitch airfoil and a delta wing with linear structural model, results are validate against a time accurate CFD-FEM code. The results show the model is able to replicate the aerodynamic forces and flow fields with sufficient accuracy while requiring a fraction of CFD cost.

Model order reduction for parameterized nonlinear evolution equations

Otto-von-Guericke-Universität Magdeburg, Fakultät für Mathematik, Dissertation, 2016