Liam Comerford

Compressive sensing based power spectrum estimation from incomplete records by utilizing an adaptive basis

A compressive sensing (CS) based approach is developed in conjunction with an adaptive basis reweighting procedure for stochastic process power spectrum estimation. In particular, the problem of sampling gaps in stochastic process records, occurring for reasons such as sensor failures, data corruption, and bandwidth limitations, is addressed. Specifically, due to the fact that many stochastic process records such as wind, sea wave and earthquake excitations can be represented with relative sparsity in the frequency domain, a CS framework can be applied for power spectrum estimation. To this aim, an ensemble of stochastic process realizations is often assumed to be available. Relying on this attribute an adaptive data mining procedure is introduced to modify harmonic basis coefficients, vastly improving on standard CS reconstructions. The procedure is shown to perform well with stationary and non-stationary processes even with up to 75% missing data. Several numerical examples demonstrate the effectiveness of the approach when applied to noisy, gappy signals.

Uncertainty Quantification of Power Spectrum and Spectral Moments Estimates Subject to Missing Data

AbstractIn this paper, the challenge of quantifying the uncertainty in stochastic process spectral estimates based on realizations with missing data is addressed. Specifically, relying on relativel...

An artificial neural network based approach for power spectrum estimation and simulation of stochastic processes subject to missing data

An artificial neural network (ANN) approach is presented as a possible solution to overcoming the problems associated with missing data in spectral analysis and/or simulation of stochastic processes. By using an ANN to capture patterns present in the available data, gaps can then be filled or entirely new processes generated. A feed-forward ANN is used with ordered inputs and Gaussian white noise to represent missing data during learning. The solution is broadly applicable in many circumstances due to the fact that it assumes no prior knowledge of the underlying statistics of the process. Specifically, to present the method in context, this paper addresses some of the challenges associated with preparing data for environmental simulation load models (time dependent, 1-dimensional).

Compressive sensing for power spectrum estimation of multi-dimensional processes under missing data

A compressive sensing (CS) based approach is applied in conjunction with an adaptive basis re-weighting procedure for multi-dimensional stochastic process power spectrum estimation. In particular, the problem of sampling gaps in stochastic process records, occurring for reasons such as sensor failures, data corruption, and bandwidth limitations, is addressed. Specifically, due to the fact that many stochastic process records such as wind, sea wave and earthquake excitations can be represented with relative sparsity in the frequency domain, a CS framework can be applied for power spectrum estimation. By relying on signal sparsity, and the assumption that multiple records are available upon which to produce a spectral estimate, it has been shown that a re-weighted CS approach succeeds in estimating power spectra with satisfactory accuracy. Of key importance in this paper is the extension from one-dimensional vector processes to a broader class of problems involving multidimensional stochastic fields. Numerical examples demonstrate the effectiveness of the approach when records are subjected to up to 75% missing data.

Utilising Database-Driven Interactive Software to Enhance Independent Home-Study in a Flipped Classroom Setting: Going beyond Visualising Engineering Concepts to Ensuring Formative Assessment.

ABSTRACT The concept of formative assessment is considered by many to play an important role in enhancing teaching in higher engineering education. In this paper, the concept of the flipped classroom as part of a blended learning curriculum is highlighted as an ideal medium through which formative assessment practices arise. Whilst the advantages of greater interaction between students and lecturers in classes are numerous, there are often clear disadvantages associated with the independent home-study component that complements timetabled sessions in a flipped classroom setting, specifically, the popular method of replacing traditional classroom teaching with video lectures. This leads to a clear lack of assurances that the cited benefits of a flipped classroom approach are echoed in the home-study arena. Over the past three years, the authors have sought to address identified deficiencies in this area of blended learning through the development of database-driven e-learning software with the capability of introducing formative assessment practices to independent home-study. This paper maps out aspects of two specific evolving practices at separate institutions, from which guiding principles of incorporating formative assessment aspects into e-learning software are identified and highlighted in the context of independent home-study as part of a flipped classroom approach.

Structural system response and reliability analysis under imcomplete earthquake records

This work is concerned with the reliability analysis of structural systems under incomplete earthquake records. An artificial neural network approach is developed and implemented to address the problem associated with missing data in the context of evolutionary power spectra estimation of underlying non-stationary stochastic processes. The effectiveness of the proposed approach is investigated by the reliability analysis of a large structural model under simulated earthquake excitation.

On quantifying the uncertainty of stochastic process power spectrum estimates subject to missing data

The issue of quantifying the uncertainty in stochastic process power spectrum estimates based on realisations with missing data is addressed. In this regard, relying on relatively relaxed assumptions for the missing data, utilising fundamental concepts from probability theory, and resorting to Fourier and harmonic wavelets based representations of stationary and non-stationary stochastic processes, respectively, a closed-form expression is derived for the probability density function (PDF) of the power spectrum value corresponding to a specific frequency. The significance of the derived PDF relates to cases where incomplete process realisations are available for power spectrum estimation applications. In this setting, standard power spectrum estimation techniques subject to missing data typically provide with a deterministic estimate for the power spectrum. Thus, no information is provided concerning the uncertainty in the estimates. Numerical examples herein demonstrate the large extent to which any given single estimate may be unrepresentative of the target spectrum.