Ratko Grbić

Review of adaptation mechanisms for data-driven soft sensors

Abstract In this article, we review and discuss algorithms for adaptive data-driven soft sensing. In order to be able to provide a comprehensive overview of the adaptation techniques, adaptive soft sensing methods are reviewed from the perspective of machine learning theory for adaptive learning systems. In particular, the concept drift theory is exploited to classify the algorithms into three different types, which are: (i) moving windows techniques; (ii) recursive adaptation techniques; and (iii) ensemble-based methods. The most significant algorithms are described in some detail and critically reviewed in this work. We also provide a comprehensive list of publications where adaptive soft sensors were proposed and applied to practical problems. Furthermore in order to enable the comparison of different methods to standard soft sensor applications, a list of publicly available data sets for the development of data-driven soft sensors is presented.

Adaptive soft sensor for online prediction and process monitoring based on a mixture of Gaussian process models

Abstract Linear models can be inappropriate when dealing with nonlinear and multimode processes, leading to a soft sensor with poor performance. Due to time-varying process behaviour it is necessary to derive and implement some kind of adaptation mechanism in order to keep the soft sensor performance at a desired level. Therefore, an adaptation mechanism for a soft sensor based on a mixture of Gaussian process regression models is proposed in this paper. A procedure for input variable selection based on mutual information is also presented. This procedure selects the most important input variables for output variable prediction, thus simplifying model development and adaptation. Apart from online prediction of the difficult-to-measure variable, this soft sensor can be used for adaptive process monitoring. The efficiency of the proposed method is benchmarked with the commonly applied recursive PLS and recursive PCA method on the Tennessee Eastman process and two real industrial examples.

Stream water temperature prediction based on Gaussian process regression

The prediction of stream water temperature presents an interesting topic since the water temperature has a significant ecological and economical role, such as in species distribution, fishery, industry and agriculture water exploitation. The prediction of stream water temperature is usually based on appropriate mathematical model and measurements of different atmospheric factors. In this paper, a probabilistic approach to daily mean water temperature prediction is proposed. The resulting model is a combination of two Gaussian process regression models where the first model describes the long-term component of water temperature and the other model describes the short-term variations in water temperature. The proposed approach is developed even further by modeling the short-term variations with multiple Gaussian process regression models instead with a single one. Apart from that, variable selection procedure based on mutual information is presented which is suitable for input variable selection when nonlinear models for stream water prediction are developed. The proposed approach is compared with traditional modeling approaches on the measurements obtained on the Drava river in Croatia. The presented methodology can be used as a basis of the predictive tools for water resource managers.

A modification of the DIRECT method for Lipschitz global optimization for a symmetric function

In this paper, we consider a global optimization problem for a symmetric Lipschitz continuous function. An efficient modification of the well-known DIRECT (DIviding RECTangles) method called SymDIRECT is proposed for solving this problem. The method is illustrated and tested on several standard test functions. The application of this method to solving complex center-based clustering problems for the data having only one feature is particularly presented.

Three points method for searching the best least absolute deviations plane

In this paper a new method for estimation of optimal parameters of a best least absolute deviations plane is proposed, which is based on the fact that there always exists a best least absolute deviations plane passing through at least three different data points. The proposed method leads to a solution in finitely many steps. Moreover, a modification of the aforementioned method is proposed that is especially adjusted to the case of a large number of data and the need to estimate parameters in real time. Both methods are illustrated by numerical examples on the basis of simulated data and by one practical example from the field of robotics.

Data Preprocessing in Data Based Process Modeling

Abstract Abstract Important process variables which give information about the final product quality cannot often be measured by a sensor. The alternative procedure is estimation of these difficult-to-measure process variables for which it is necessary to have an appropriate process model. Process model building is based on plant data, taken from the process database. Since the quality of the built model depends heavily on the modeling data informativity, a preparatory part of modeling, in which analysis and preprocessing of available measured data are performed, is a very important step in such process modeling. The analysis and preprocessing of plant data obtained from an oil distillation process are showed in the paper. The results show that, apart from the regression method applied, selection of easy-to-measure variables which will be used in the model building and filtering of easy-to-measure variables significantly affects process model prediction capabilities.

Adaptive soft sensor for online prediction based on moving window Gaussian process regression

Very often important process variables cannot be measured online due to low sampling rate of sensors or because their values have to be obtained by laboratory analysis. In order to enable continuous process monitoring and efficient process control in such cases, soft sensors are usually used to estimate these difficult-to-measure process variables. Most industrial processes exhibit some kind of time-varying behavior. To ensure that soft sensor retains its precision, adaptation mechanism has to be implemented. In this paper adaptive soft sensor based on Gaussian Process Regression (GPR) is presented. To make GPR model training more efficient, algorithm for variable selection based on Mutual Information is proposed. Prediction capabilities of the proposed method are examined on real industrial data obtained at an oil distillation column.

A method for solving the multiple ellipses detection problem

In this paper, the multiple ellipses detection problem on the basis of a data points set coming from a number of ellipses in the plane not known in advance is considered. An ellipse is considered as a Mahalanobis circle with some positive definite matrix. A very efficient method for solving this problem is proposed. This method very successfully combines the well-known direct least squares method and the RANSAC algorithm with a realistic statistical model of multiple ellipses in the plane. The method is illustrated and tested on numerous synthetic and real-world applications. The method was also compared with other similar methods. In the case when a data points set comes from a number of ellipses with clear edges, the proposed method gives results similar to other known methods. However, when a data points set comes from a number of ellipses with noisy edges, the proposed method performs significantly better than the other methods. We should emphasize the advantage and utility of the proposed methods in a variety of applications such as: medical image analysis and ultrasound image segmentation. HighlightsRANSAC-based method for solving the multiple ellipses detection problem is proposed.The proposed method shows high efficiency.The method has the potential to solve real time applications.Data is considered as a realization of a realistic statistical model.Pseudocodes of all proposed algorithms are provided.

Approximating surfaces by the moving least absolute deviations method

In this paper we are going to consider the problem of global data approximation on the basis of data containing outliers. For that purpose a new method entitled the moving least absolute deviations method is proposed. In the region of data in the network of knots weighted least absolute deviations local planes are constructed by means of which a global approximant is defined. The method is tested on the well known Frankes function. An application in gridding of sonar data is also shown.

Picture quality meter — No-reference video artifact detection tool

The demand for a wide range of video and multimedia applications is growing extremely fast in recent years. In such a diverse usage of digital video, there is a need for reliable video quality assessment by different parties involved in video content delivery to end users. While video quality assessment methods are gaining quite considerable attention in scientific research, practical tools for video quality assessment are still rare. This article presents Picture Quality Meter, an application for real-time no-reference video artifact detection that can be used in different scenarios like video equipment testing, network monitoring, video hardware/software development and scientific research. The application has some unique features allowing user customization in terms of used no-reference methods, giving the user deeper insight into the obtained results and easier usage when dealing with in field video equipment testing.