Dynamic Prediction of Project Success Using Artificial Intelligence

The purpose of construction management is to successfully accomplish projects, which requires a continuous monitoring and control procedure. To dynamically predict project success, this research proposes an evolutionary project success prediction model (EPSPM). The model is developed based on a hybrid approach that fuses genetic algorithms (GAs), fuzzy logic (FL), and neural networks (NNs). In EPSPM, GAs are primarily used for optimization, FL for approximate reasoning, and NNs for input-output mapping. Furthermore, the model integrates the process of continuous assessment of project performance to dynamically select factors that influence project success. The validation results show that the proposed EPSPM, driven by a hybrid artificial intelligence technique, could be used as an intelligent decision support system, for project managers, to control projects in a real time base.     

Techniques for Predicting Cracking Pattern of Masonry Wallet Using Artificial Neural Networks and Cellular Automata

This paper introduces innovative artificial intelligent techniques for directly predicting the cracking patterns of masonry wallets, subjected to vertical loading. The von Neumann neighborhood model and the Moore neighborhood model of cellular automata (CA) are used to establish the CA numerical model for masonry wallets. Two new methods—(1) the modified initial value method and (2) the virtual wall panel method—that assist the CA model are introduced to describe the property of masonry wallets. For practical purposes, techniques for the analysis of wallets whose bed courses have different angles with the horizontal bottom edges are also introduced. In this study, two criteria are used to match zone similarity between a “base wallet” and any new “unseen” wallets. This zone similarity information is used to predict the cracks in unseen wallets. This study also uses a back-propagation neural network for predicting the cracking pattern of a wallet based on the proposed CA model of the wallet and some data of recorded cracking at zones. These techniques, once validated on a number of unseen wallets, can provide practical innovative tool for analyzing structural behavior and also help to reduce the number of expensive laboratory test samples.     

Dynamic Regression Models for Prediction of Construction Costs

Accurate prediction of construction costs in the market is essential to effectively estimate costs for construction projects. In the construction industry, cost indexes that are reported in series are often used to explain the change of construction costs. By tracking the trend of such quantitative contemporaneous cost index and making frequent and regular forecasts of the future values of the index, one can develop a deeper understanding of prices of resources used for construction. Incorporating such an understanding and prediction into estimating will help practitioners manage construction costs. This paper proposes two dynamic regression models for the prediction of construction cost index. Comparison of the proposed models with the existing methods proves that the new models provide several advantages and improvements.     

Using Complex Permittivity and Artificial Neural Networks for Contaminant Prediction

The use of the measured complex permittivity of electrolyte solutions for predicting ionic species and concentration is investigated. Four artificial neural networks (ANNs) are created using a database containing permittivities (at 1.0, 1.5, and 2.0 GHz) and loss factors (at 0.3, 1.5, and 3.0 GHz) of 12 aqueous salts at various concentrations. The first ANN correctly identifies cationic species in 83% of the samples and distinguishes between pure water and electrolyte solutions with 100% accuracy. The second ANN predicts cationic concentrations with a RMS error of 190 mg/L for the range of concentrations examined (0–3,910 mg/L) and explains 90% of the variability in these data. The third ANN correctly identifies 98% of the anionic species in samples and accurately distinguishes between pure water and anion-containing solutions. The last ANN predicts anionic concentrations with a RMS error of 164 mg/L for the range of concentrations examined (0–5,654 mg/L) with an r2 of nearly 98%.     

Parameter Estimation in Groundwater Hydrology Using Artificial Neural Networks

The capability of artificial neural networks to act as universal function approximators has been traditionally used to model problems in which the relation between dependent and independent variables is poorly understood. In this paper, the capability of an artificial neural network to provide a data-driven approximation of the explicit relation between transmissivity and hydraulic head as described by the groundwater flow equation is demonstrated. Techniques are applied to determine the optimal number of nodes and training patterns needed for a neural network to approximate groundwater parameters for a simulated groundwater modeling case study. Furthermore, the paper explains how such an approximation can be used for the purpose of parameter estimation in groundwater hydrology.     

Effectiveness of Different Artificial Neural Network Training Algorithms in Predicting Protozoa Risks in Surface Waters

A neural network approach was employed to relate risky Cryptosporidium and Giardia concentrations with other biological, chemical and physical parameters in surface water. A set of drinking water samples was classified as “risky” and “nonrisky” based on the concentrations of full and empty oocysts, and cycsts of Cryptosporidium and Giardia, respectively. Given the constraints associated with collecting large sets of microbial data, the study was aimed at identifying an effective training algorithm that would maximize the performance of a neural network model working with a relatively small dataset. A number of algorithms for training neural networks, including gradient search with first- and second-order partial derivatives, and genetic search were used and compared. Results showed that genetic algorithm based neural network training consistently provided better results compared to other training methods.     

Short-Term Traffic Prediction on Different Types of Roads with Genetically Designed Regression and Time Delay Neural Network Models

Research for advanced traveler information systems (ATIS) has been focused on urban roads. However, research for short-term traffic prediction on all categories of highways is needed, as highway agencies expect to implement intelligent transportation systems across their jurisdictions. In this study, genetic algorithms were used to design time delay neural network (TDNN) models as well as locally weighted regression models to predict short-term traffic for six rural roads from Alberta, Canada. These roads are from various trip-pattern groups and functional classes. Refined TDNN models developed in this study can limit most average errors less than 10% for all study roads. Refined regression models show even higher accuracy. Average errors for the refined regression models are less than 2% for roads with stable patterns. Even for roads with unstable patterns, average errors are below 4%, and the 95th percentile errors are less than 7%. It is believed that such accurate predictions would be useful for highway agencies to implement statewide ATIS.     

Artificial Intelligence Techniques as Detection Tests for the Identification of Shifts in Hydrometric Data

This paper presents the development of tests based on one artificial intelligence technique, the Kohonen neural network, for the detection of shifts in hydrometric data. Two new Kohonen-based detection tests are developed, the classification and mapping tests, and their performance is compared with that of well-known conventional detection tests. The efficacy of the tests is demonstrated with synthetic data, for which all the statistical properties and induced shifts are known. These synthetic data are designed to replicate hydrometric data such as annual mean and maximum streamflow. The results show that all tests, conventional and Kohonen based, may be considered equally reliable. However, no one test should be used alone because all generate false diagnostics under different circumstances. Within a decision support environment, a pool of tests may be used to confirm or complement one another depending on their known strengths and weaknesses. The Kohonen-based detection tests also perform well when applied to multivariate cases (i.e., testing more than one data sequence at a time), and their performance for multivariate cases is better than that for the univariate cases.     

Prediction of Soil Composition from CPT Data Using General Regression Neural Network

Soil type is typically inferred from the information collected during a cone penetration test (CPT) using one of the many available soil classification methods. In this study, a general regression neural network (GRNN) was developed for predicting soil composition from CPT data. Measured values of cone resistance and sleeve friction obtained from CPT soundings, together with grain-size distribution results of soil samples retrieved from adjacent standard penetration test boreholes, were used to train and test the network. The trained GRNN model was tested by presenting it with new, previously unseen CPT data, and the model predictions were compared with the reference particle-size distribution and the results of two existing CPT soil classification methods. The profiles of soil composition estimated by the GRNN generally compare very well with the actual grain-size distribution profiles, and overall the neural network had an 86% success rate at classifying soils as coarse grained or fine grained.     

Comparative Study of SVMs and ANNs in Aquifer Water Level Prediction

In this research, a data-driven modeling approach, support vector machines (SVMs), is compared to artificial neural networks (ANNs) for predicting transient groundwater levels in a complex groundwater system under variable pumping and weather conditions. Various prediction horizons were used, including daily, weekly, biweekly, monthly, and bimonthly prediction horizons. It was found that even though modeling performance (in terms of prediction accuracy and generalization) for both approaches was generally comparable, SVM outperformed ANN particularly for longer prediction horizons when fewer data events were available for model development. In other words, SVM has the potential to be a useful and practical tool for cases where less measured data are available for future prediction. The study also showed high consistency between the training and testing phases of modeling when using SVM compared to ANN. While for the proposed SVM model the relative error of mean square error increased by an average of 42% from the training phase to testing the phase, the corresponding testing error of the ANN model raised by approximately seven times the training error.     

Case Study: Finite Element Method and Artificial Neural Network Models for Flow through Jeziorsko Earthfill Dam in Poland

A finite element method (FEM) and an artificial neural network (ANN) model were developed to simulate flow through Jeziorsko earthfill dam in Poland. The developed FEM is capable of simulating two-dimensional unsteady and nonuniform flow through a nonhomogenous and anisotropic saturated and unsaturated porous body of an earthfill dam. For Jeziorsko dam, the FEM model had 5,497 triangular elements and 3,010 nodes, with the FEM network being made denser in the dam body and in the neighborhood of the drainage ditches. The ANN model developed for Jeziorsko dam was a feedforward three layer network employing the sigmoid function as an activator and the back-propagation algorithm for the network learning. The water levels on the upstream and downstream sides of the dam were input variables and the water levels in the piezometers were the target outputs in the ANN model. The two models were calibrated and verified using the piezometer data collected on a section of the Jeziorsko dam. The water levels computed by the models satisfactorily compared with those measured by the piezometers. The model results also revealed that the ANN model performed as good as and in some cases better than the FEM model. This case study offers insight into the adequacy of ANN as well as its competitiveness against FEM for predicting seepage through an earthfill dam body.     

Application of Artificial Intelligence to Estimate Daily Pan Evaporation Using Available and Estimated Climatic Data in the Khozestan Province (South Western Iran)

Estimation of evaporation, a major component of the hydrologic cycle, is required for a variety of purposes in water resources development and management. This paper investigates the abilities of genetic programming (GP) to improve the accuracy of daily evaporation estimation. In the first part of the study, different GP models, comprising various combinations of daily climatic variables, namely, air temperature, sunshine hours, wind speed, and relative humidity, were developed to evaluate the degree of the effect of each variable on daily pan evaporation. A dynamic modeling of evaporation was also performed, with the current climatic variables and one of the previous variables, to evaluate the effect of their time series on evaporation. In the second part of the study, the estimated solar radiation data were used as input vectors instead of recorded sunshine values. Statistics such as correlation coefficient (R), root mean square error (RMSE), coefficient of residual mass (CRM) and scatter index (SI) were used to measure the performance of models. Tthe dynamic model approach was shown to give the best results with relatively fewer errors and higher correlations. To assess the ability of GP relative to the neuro-fuzzy (NF) and artificial neural networks (ANN), several NF and ANN models were developed by using the same data set. The obtained results showed the superiority of GP to the NF and ANN approaches. The Stephen-Stewart and Christiansen methods were also considered for comparison. The results indicated that the proposed GP model performed quite well in modeling evaporation processes from the available climatic data. The results also showed that the estimated solar radiation data can be applied successfully instead of the recorded sunshine data.     

人工神经网络在钢铁材料力学性能预测方面的应用

人工神经网络模型特别适用于非线性系统。具有较好的学习精度和概括能力。已成功应用于钢铁材料力学性能的预测。使用人工神经网络模型，通过输入合金元素、组织、生产工艺参数可预测钢铁材料的抗拉强度、延伸率、韧性、疲劳和蠕变性能。概要叙述了人工神经网络在预测板材、球墨铸铁的常温力学性能，合金结构钢的淬透性。高速钢、不锈耐热钢的热强度以及微合金钢热扭转性能方面的应用。     

增量神经网络模型预报100 t电弧炉终点碳、磷和温度的应用

结合增量模型和神经网络模型的优点,提出增量神经网络模型,该模型特点为只注重系统输入量和输出量的变化,系统输入与输出增量的映射关系通过网络很快形成,网络结构简单.以废钢、铁水、装料制度、通电时间、吨钢氧耗和电耗相对于参考炉均值的增量为输入节点,对冶炼钢水终点温度和碳、磷进行预报.结果表明,当钢水终点温度和碳、磷含量的控制精度分别在±10 ℃,±0.02%和±0.004%时,预报值命中率分别为93%,75%和86%.