Figure 1 - available via license: CC BY
Content may be subject to copyright.
Source publication
In refinery plants key process variables, like contents of process stream and various fuel properties, need to be continuously monitored using adequate on-line measuring devices. Such measuring devices are often unavailable or malfunction and, hence, laboratory assays which are irregular and time consuming and therefore not suitable for process con...
Contexts in source publication
Context 1
... fractionation reformate plant with the variables used for soft sensor development is given in Figure 1. Reformate enters into column C-1 where the light reformate is separated from the mixture of heavy reformate and benzene fraction. ...
Context 2
... in the case of the linear model developed on generated data, the MARSpline model achieved approximately same statistical values in all three subsets, although somewhat better than the linear model. Very good matching of the MARSpline model and experimental data on the validation data set can be seen in Figures 9 and 10. ...
Context 3
... parameters of the neural network models developed on the small data set are shown in Table 7. High values of correlation coefficients and small errors point out that the model very well describes the actual data. Such good matching with minor deviations is also observed in Figures 11 and 12. ...
Context 4
... statistical parameters with almost the same values for the estimation and validation data sets show high model accuracy, which is better than the MLP model developed on the small data set. From Figure 13, it can be seen that deviations from the direction y = x are minimal on the entire dataset. In Figure 14 for the histogram, it is clear that most of the errors lie between -0.1 and +0.1 vol. ...
Similar publications
This paper is dedicated to development of mathematical models for polynomial spline curve formation given extreme vector derivatives. This theoretical problem is raised in the view of a wide variety of theoretical and practical problems considering motion of physical objects along certain trajectories with predetermined laws of variation of speed,...
Citations
... Besides Inline Analyzers and Laboratory Analysis, Virtual Online analyzers (VOA) (Soft Sensors or Virtual Online Analyzers) are used. Most of them are based on multivariate regression models [6,7] or artificial intelligence techniques [8,9,10]. ...
... where b1-b3 are influence coefficients of single (lone) factors; b12, b13, b23 are interference ratios of factor pairs; b4-b6 are influence coefficients of factor squares; x0-x3 are encoded factors; φ is biased factor square values; y is an output (result) parameter (variable). The coefficients of a regression equation (6) After the test of significance of b-coefficients and insignificant coefficients eliminated from the model, a regression equation (6) was rearranged into the standard form: ...
Mathematical and simulation modeling methods are applied when defining parameters of effective operation of oil fractionation. There is a review of software packages of process modeling used for refining ventures in the study. Multisphere model of oil fractionation with a low capacity applied is offered in the paper. The designed model allows taking into account constructive peculiarities of the equipment of small capacity units. The flow diagram of the model calculation algorithm with the additional evaluator of petroleum products quality indices is given by the authors. A design matrix for a three-factor experiment involving a multisphere model is obtained. An orthogonal central composite design has been applied to improve the accuracy of the model and to consider a non-linear parameter link. The MatLab software was chosen as an enabler. The criterion was the energy cost of separation. Factors affecting the criterion are flow rate of superheated water steam, reflux rate, potential light fractions in the crude/feed. The adequate obtained regression quadratic dependency towards the initial values by Fisher’s criterion has been assessed. The average relative error was 0.21 percent. The obtained statistical model can be applied for solving an optimization problem by the criterion of power consumption within a variable fractional composition of the raw material. Quality indices of oil products (gasoline fraction and diesel oil) serve as constraints. The offered statistical model applied at oil fractionation low-power facilities promotes the efficiency of plant operation improvement and the efficiency of obtaining information about the quality of oil products.
... The bootstrap (BS) [21] and jackknife (JK) [22] methods are widely used in the construction of SSs (or models) to estimate the quality indices of end products under the conditions of small samples. These methods make it possible to estimate the SS parameters via the construction of new samples from the initial dataset. ...
... After the type of distribution densities and is determined, their intersection area is calculated. In the case of intersection of with at point c (assuming that and have lognormal distribution functions), S can be calculated as follows: (21) where c is the point of intersection and and After the transformation, we have (22) where and are the mean values of and and and are their standard deviations, respectively. ...
... In the case when the technological mode data are completely absent for some stationary states, the methods described above are not applicable.The use of a low-informative training segment of data leads to the ineffectiveness of the algorithms of building process models, as small training samples do not have enough information to obtain an SS that provides an acceptable assessment of the quality indicators of the output product. To solve the problem of building SS under conditions of small sample size, bootstrap analysis [4], as well as various methods based on neural networks (NN) and their various modifications and combinations, such as NN with radial base functions and deep convolution NN with bootstrap [5], were widely used. The use of bootstrap analysis reduces the risk of model relearning by replacing the training set with bootstrap samples. ...
The paper discusses the development of a soft sensor (SS) for an MTBE plant in the cases when the training sample either is small or does not comprise the whole of quality range because of process plant non-stationarity as well as the difficulty and high cost of retrieving more information. In order to build a soft sensor that provides higher accuracy in estimating the quality of the output product, an algorithm for extending the initial training sample using a rigorous model of the distillation column of the methyl-tert-butyl ether production under condition of exactly unknown values of the Murphree trays efficiency has been proposed. The resulting SS enables 40% improvement of product quality prediction.
... Certain process variables such as exhaust gas concentrations are difficult to measure in real time and, therefore, they are classically monitored by offline laboratory analysis (Andrijić et al. 2018;Lin et al. 2007). Composition sensors with small real time sampling frequencies are not commonly encountered in industrial process plants, due to high operational costs of specific hard sensors. ...
Soft sensors with real time prediction capabilities appear as a profitable solution for hard-to-measure variables whenever hard sensors are difficult to apply or subjected to high operational costs. Nonetheless, the use of soft sensors within industrial applications is still not widespread because of the systematic accuracy issues that can be introduced with process plant deviations from nominal operation states. Soft sensor models need to be constantly updated to avoid degradation of their prediction potential. This study presents an innovative view on a well-known artificial neural network (ANN) calibration method by developing a generic online calibration tool that can be used in independent data-driven soft sensors based on ANN multi-layer perceptron (MLP) models. The maintenance framework has been fully tested in a semi-industrial boiler plant to predict real time pollutant emission levels, presenting recalibration time responses up to 1 min, overall r2 performance above 80% and an intuitive human–machine-interface.
Bez analitičke kemije nemoguće je zamisliti napredak mnogih područja ljudske djelatnosti. Spomenuti značaj ove grane kemije jedan je od glavnih razloga uspješnog opstanka i kontinuiranog razvoja Zavoda za analitičku kemiju unutar jednog tehničkog fakulteta, kao što je današnji Fakultet kemijskog inženjerstva i tehnologije (FKIT) Sveučilišta u Zagrebu. Dinamičnost razvoja analitičke kemije bila je neprestani poticaj nastavnicima Zavoda za prilagođavanjem svjetskim trendovima i sustavnim unaprjeđenjem nastavnog programa i istraživačkog rada.
Početak nastavnog, znanstvenog, ali i stručnog djelovanja Zavoda temeljio se na klasičnim metodama analize, no potaknuta razvojem moderne analitičke opreme ubrzo se intenzivirala primjena brojnih kemijskih i fizikalno-kemijskih metoda analize, od kojih je danas najzastupljenija kromatografija. Naime, konac dvadesetog stoljeća ubrzanim je razvojem instrumentalnih tehnika, uvođenjem računala i činjenicom da su podatci analitičkih mjerenja postali temelj odlučivanja u medicini, zaštiti okoliša, razvoju materijala te osiguravanju kvalitete procesa i proizvoda, dao novi uzlet analitičkoj kemiji, a osobit procvat doživjela je kromatografija kao idealna tehnika za višekomponentnu analizu analita u tragovima.
