A data-driven study for evaluating fineness of cement by various predictors
- PDF / 993,558 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 53 Downloads / 120 Views
ORIGINAL ARTICLE
A data-driven study for evaluating fineness of cement by various predictors Bulent Tutmez
Received: 14 February 2014 / Accepted: 23 June 2014 Ó Springer-Verlag Berlin Heidelberg 2014
Abstract Modelling relationships among cement and concrete parameters from different perspectives is preferred due to its practical importance. The relationship between chemical ingredients and specific surface area which addresses fineness of cement were appraised via three predictors: robust regression (RR), support vector regression (SVR) and multi-layer perception (MLP). The main motivation of the study was to give a comparative assessment with sparse data based on accuracy of the models. In addition to accuracy, smoothing level of the estimations was also considered and the performances of three models were compared with the former practices. The experimental studies showed that the SVR model performs better than the rest of the models for identifying the relationships. The potentials of the MLP and the RR models have also been discussed. Keywords Fineness of cement Regression Support vector machine Neural network
1 Introduction Cement is a general nomenclature and is known the entire world over as Portland cement (PC). It has critical importance for construction and building; therefore, it also must have certain qualities. While it is important to have the correct proportions of calcium, silicon and aluminum, the overall chemical composition and structure of the individual raw ingredients can vary considerably. Some B. Tutmez (&) School of Engineering, Inonu University, 44280 Malatya, Turkey e-mail: [email protected]
natural rocks contain these ingredients like CaO, SiO2 and Al2O3 in approximately right proportions [8]. One of the important steps in the manufacture of PC is to combine a variety of raw ingredients so that the resulting cement will have the desired chemical composition [22]. The size of a cement particle has an important effect on the rate at which it will hydrate when exposes to water. A better parameter for describing the fineness of the cement is the specific surface area (Blaine), because most of the surface area comes from the smallest particles. Thus, the chemical composition and Blaine affect the workability of the cement at a given water/cement ratio and the degree of hydration and strength development over long periods of time, among other factors. There is a connection between the chemical composition and the specific surface area and evaluating this relationship maintains its importance [24]. In recent years, some studies have been conducted for modelling or optimizing cement and concrete parameters from different perspectives. In the first perspective, the problem was considered as a stochastic problem and some regression-based tools such as least squares and ridge regression were applied [11, 23, 25]. In the second perspective, the problem was handled by computational intelligence-based approaches [15, 16, 27, 30]. In general, the former studies mainly focused on pr
Data Loading...
