Search

This note investigates the potential of support vector machines based regression approach to model the static pile capacity from dynamic stress-wave data. A data set of 105 prestressed precast high strength concrete spun pipe piles is used. Radial basis function and polynomial kernel based support vector machines were used to model the total pile capacity and results were compared with a generalized regression neural network approach. A total of 81 data set were used to train, whereas the remaining 24 data sets were used to test the created model. A correlation coefficient value of 0.977 was achieved by generalized regression neural network in comparison to values of 0.967 and 0.964 achieved by radial basis function and polynomial kernel based support vector machines, respectively. Results suggest an improved performance by generalized regression neural network based approach in comparison to support vector machines but polynomial kernel based support vector machines provide a linear relationship to predict total pile capacity using stress-wave data. CE Database subject headings: Piles; Neural networks; Vector analysis; Stress wave.

Linking of the activities in a critical path method schedule with the corresponding elements of a three-dimensional (3D) model makes the project sequence easier to understand. Although some commercial tools allow the planner to build a 4D model and create graphical simulation of the construction process, it still lacks features like generation and manipulation of a 4D model within a single environment. This paper presents a geographic information systems (GIS) based methodology for scheduling as an alternative to the existing 4D computer aided design tools. A methodology to build a 3D model and link it with the construction schedule using several in-house scripts written in GIS environment is discussed. It allows the planner to understand the construction schedule quickly by linking the activities of the schedule with the corresponding 3D components as well as to visualize a buildable schedule on a computer screen. The proposed methodology utilizes the dynamic linkage between the activities in the schedule and corresponding 3D components, thus, making it possible to detect the incompleteness and logical errors in the schedule sequence. The ability of GIS to maintain spatial data (i.e., 3D components corresponding to each activity) in separate themes, which can be superimposed spatially, is utilized. The related nonspatial information (like schedule, material type and quantity, safety and quality control recommendations, etc.) are stored in the attribute tables of corresponding components, which can be extracted from the database maintained within the GIS itself. CE Database subject headings: Geographic information systems; Scheduling; Construction management; Project management; Information systems.

An artificial neural network (ANN) and support vector machines (SVMs) were employed for estimating the removal efficiency of settling basins in canals. The performance of ANN and SVMs was tested using the data from an earlier study carried out by Ranga Raju et al. As compared with the Ranga Raju et al. relationship, the correlation coefficient of ANN as well as SVMs improved from 0.77 to 0.9854 and 0.9853, respectively; whereas the root mean squared error values decreased from 50.66 to 5.712 and 5.7366, respectively. Between SVMs and ANN, SVMs’ performance was found to be better due to its use of the principle of structural risk minimization in formulating cost functions and the use of quadratic programming during model optimization. These advantages led to a unique optimal solution as compared to conventional neural network models. [ABSTRACT FROM AUTHOR]

The article comments on the article "Application of Support Vector Machines in Assessing Conceptual Cost Estimates," by Sung-Hoon An, U-Yeol Park, Kyung-in Kang, Moon-Young Cho and Hun-Hee Cho, published in the July/August 2007 issue. It presents observations about issues related to the design of support vector machine (SVM) and the methods used in accuracy assessment. The authors question the accuracy of claims of Sung-Hoon et al., that a kernel function satisfying the nonlinear support vector classifier is known as the function satisfying Mercer's condition. The limited evaluation of the development of SVM model using only one statistical performance evaluation criterion is also questioned by the authors.

A commentary on the article "Construction Scheduling and Progress Control Using Geographical Information Systems," by Stephen E. Poku and David Arditi, is presented. The authors address the failure of traditional scheduling and progress control techniques to offer information about the spatial aspects of a construction project. They recommend the use of backend programming to make a proposed system more user friendly.