Andriy Ivannikov

Online mass flow prediction in CFB boilers with explicit detection of sudden concept drift

Fuel feeding and inhomogeneity of fuel typically cause fluctuations in the circulating fluidized bed (CFB) process. If control systems fail to compensate the fluctuations, the whole plant will suffer from dynamics that is reinforced by the closed-loop controls. This phenomenon causes reducing efficiency and the lifetime of process components. In this paper we address the problem of online mass flow prediction, which is a part of control. Particularly, we consider the problem of learning an accurate predictor with explicit detection of abrupt concept drift and noise handling mechanisms. We emphasize the importance of having domain knowledge concerning the considered case and constructing the ground truth for facilitating the quantitative evaluation of different approaches. We demonstrate the performance of change detection methods and show their effect on the accuracy of the online mass flow prediction with real datasets collected from the experimental laboratory-scale CFB boiler.

Extraction of the mismatch negativity elicited by sound duration decrements: A comparison of three procedures

This study focuses on comparison of procedures for extracting the brain event-related potentials (ERPs) - brain responses to stimuli recorded using electroencephalography (EEG). These responses are used to study how the synchronization of brain electrical responses is associated with cognition such as how the brain detects changes in the auditory world. One such event-related response to auditory change is called mismatch negativity (MMN). It is typically observed by computing a difference wave between ERPs elicited by a frequently repeated sound and ERPs elicited by an infrequently occurring sound which differs from the repeated sounds. Fast and reliable extraction of the ERPs, such as the genuine MMN, is an important focus of studies devoted to basic cognitive brain research. In this study, we compared three procedures for extraction of the MMN elicited by infrequent duration decrements in auditory sound stimulation. These were the conventional difference wave (DW) with average standard sweep, optimal digital filtering (ODF), and recently proposed independent component analysis (ICA) decomposition procedures. The statistical comparison was made in a group of 12 healthy adults aged 23-30 years. The MMN was elicited in a passive oddball protocol presenting an auditory stimulation consisting of frequent tones of 600Hz of 100ms duration each (standard stimuli). Infrequently, one of the tones was shortened to 75, 50, or 30ms (deviant stimuli). The ICA decomposition procedure, similarly to the DW procedure with average standard sweep, extracted a cleaner MMN compared to the ODF procedure. Both procedures extracted the MMN, whose amplitude and latency characteristics concur with substantial number of publications in contrast to the ODF procedure.

Handling outliers and concept drift in online mass flow prediction in CFB boilers

In this paper we consider an application of data mining technology to the analysis of time series data from a pilot circulating fluidized bed (CFB) reactor. We focus on the problem of the online mass prediction in CFB boilers. We present a framework based on switching regression models depending on perceived changes in the data. We analyze three alternatives for change detection. Additionally, a noise canceling and a state determination and windowing mechanisms are used for improving the robustness of online prediction. We validate our ideas on real data collected from the pilot CFB boiler.

ERP denoising in multichannel EEG data using contrasts between signal and noise subspaces

In this paper, a new method intended for ERP denoising in multichannel EEG data is discussed. The denoising is done by separating ERP/noise subspaces in multidimensional EEG data by a linear transformation and the following dimension reduction by ignoring noise components during inverse transformation. The separation matrix is found based on the assumption that ERP sources are deterministic for all repetitions of the same type of stimulus within the experiment, while the other noise sources do not obey the determinancy property. A detailed derivation of the technique is given together with the analysis of the results of its application to a real high-density EEG data set. The interpretation of the results and the performance of the proposed method under conditions, when the basic assumptions are violated - e.g. the problem is underdetermined - are also discussed. Moreover, we study how the factors of the number of channels and trials used by the method influence the effectiveness of ERP/noise subspaces separation. In addition, we explore also the impact of different data resampling strategies on the performance of the considered algorithm. The results can help in determining the optimal parameters of the equipment/methods used to elicit and reliably estimate ERPs.

Non-smooth SOR for L1-Fitting: Convergence Study and Discussion of Related Issues

In this article, the denoising of smooth (H1-regular) images is considered. To reach this objective, we introduce a simple and highly efficient over-relaxation technique for solving the convex, non-smooth optimization problems resulting from the denoising formulation. We describe the algorithm, discuss its convergence and present the results of numerical experiments, which validate the methods under consideration with respect to both efficiency and denoising capability. Several issues concerning the convergence of an Uzawa algorithm for the solution of the same problem are also discussed.

Extraction of ERP from EEG data

In this article, a simple but novel technique for extracting a linear subspace related to event related potentials (ERPs) from ElectroEncephaloGraphy (EEG) data is introduced. The technique consists of a sequence of basic linear operations applied to multidimensional EEG data in a problem-specific manner. The derivation of the proposed technique is given and results with real data are described together with overall conclusions.

Online Mass Flow Prediction in CFB Boilers

Fuel feeding and inhomogeneity of fuel typically cause process fluctuations in the circulating fluidized bed (CFB) process. If control systems fail to compensate for the fluctuations, the whole plant will suffer from fluctuations that are reinforced by the closed-loop controls. This phenomenon causes a reduction of efficiency and lifetime of process components. Therefore, domain experts are interested in developing tools and techniques for getting better understanding of underlying processes and their mutual dependencies in CFB boilers. In this paper we consider an application of data mining technology to the analysis of time series data from a pilot CFB reactor. Namely, we present a rather simple and intuitive approach for online mass flow prediction in CFB boilers. This approach is based on learning and switching regression models. Additionally, noise canceling, and windowing mechanisms are used for improving the robustness of online prediction. We validate our approach with a set of simulation experiments with real data collected from the pilot CFB boiler.

Spatial weighted averaging for ERP denoising in EEG data

In the present paper we intend to improve the practical accuracy of ERP denoising methods proposed in earlier research by allowing them to take into account possible violations of the underlying assumptions, which often take place in practice. Here we consider ERP denoising approaches operating within the framework of the linear instantaneous mixing model that consist three steps: (1) forward linear transformation, (2) identification of the components related to signal and noise subspaces, (3) inverse transformation during which the components that belong to the noise subspace are disregarded, i.e. dimension reduction in the component space. The separation matrix is found based on problem-specific assumptions formalized in terms of the second-order statistics. The subspace separation problem is concerned rather than the source separation. For the purpose of increasing the accuracy of spatial separation of ERP and noise sources we propose a spatial weighted averaging method analogous to weighted averaging technique developed for single channel use, which takes into account variable variances of the sources over trials.

A robust approach to ERP denoising

The purpose of presented study is to explore possibilities to increase the robustness and improve the performance of the spatial ERP denoising methods proposed in earlier research. The quality of the subspace separation solution may easily be degraded essentially, if the underlying assumptions become noticeably violated, which is a normal situation in practice. The distortions to the results of a separation are caused by non-zero sample signal-noise and noise-noise correlations, which are indistinguishable from the variances of the signal and noise in the framework of the second-order statistical information exploited by the discussed methods. Therefore, in the research reported in this article we concentrate our efforts on finding the means that allow to reduce the erroneous influence of undesirable correlations on the performance of the discussed denoising methods.

Separation of ERP and Noise Subspaces in EEG Data without Whitening

In this article, a method for separating linear subspaces of time-locked brain responses and other noise sources in multichannel electroencephalography data is proposed. The components related to time-locked and noise subspaces are distinguished by method based on different behavior they experience after traditional averaging. The actual separation of the two subspaces is performed without whitening by maximizing/minimizing the same criterion. The detailed derivation of the method is given, and the results of the methods application to simulated and real EEG datasets are studied. The possibilities of improving the results are also discussed.