Search

In this study, 3-layer and 4-layer neural networks are designed to determine the rutting performance of asphalt concrete. A total of 519 sets of processed data obtained from mix design information and laboratory tests are used for developing this NN model. Finally, the NN selected has 11 neurons in each hidden layer, whereas the output layer uses a total of 7 neurons. Using a total of 21 inputs, the developed model produces outputs (rut depths) at 7 different cycles. The time series of deformation recorded over 8000 cycles are determined by an interpolation with piecewise linear elements, using these few outputs. Preprocessing and principal component analyses are applied, and the network trained using the Levenberg-Marquardt algorithm. Using randomly generated weight factors to initialize the training algorithm, histograms are compiled and outputs estimated using statistical estimators. An excellent agreement is observed between test data and simulations. It is believed that the developed NN design procedure will be a useful tool in the study of pavement design and wear. [ABSTRACT FROM AUTHOR]

This paper presented a reliability-based approach for the modeling of the deterioration and service life of concrete structures taking into account the uncertainties in the physical modeling, and variability of the material and structural parameters governing the chloride penetration into concrete and corrosion of the reinforcing steel. It illustrated the application of a probabilistic approach for the uncertainty modeling and prediction of chloride contamination of concrete and reinforcement corrosion in bridge structures that are subjected to the application of deicing salts during winter. The proposed probabilistic model provided very good predictions of the level of chloride contamination at different depths as well as the extent of corrosion of the reinforcing steel in the top mat of a deteriorated reinforced concrete bridge deck. The application of such an approach is required in the assessment of safety and serviceability of deteriorating concrete structures in order to ensure that the probability of failure is kept at an acceptable level. A reliability-based prediction of the service life of deteriorating concrete structures provides a rational decision support tool at both the initial design stage and during the operation and maintenance stage. The implementation of such an approach allows adequate control of the safety and serviceability of the structure throughout its service life and yields low life cycle cost. [ABSTRACT FROM AUTHOR]

The study conducted showed that the definitions of the fuzzy sets associated to the input variables of a given fuzzy signal controller must be taken very carefully as they affect both the actual controller response and the traffic performance. Although not studied here, it is possible to infer from the results that if the controller and traffic operations are sensitive to small variations in the input variables' fuzzy sets, they should also be very much influenced by modifications to the output variable fuzzy sets. For the situation analysed in this work, it was found that increasing the fuzzy sets limits causes general improvement for the traffic performance at the Minor Street. These, as well as other numerical results are of course dependent on the characteristics of the intersection and control strategy in question. Thus, in order to allow for the traffic control objectives at a given intersection to be well captured by the fuzzy signal controller, special care must be taken in defining the fuzzy sets linked to all control linguistic variables. A very straightforward method for capturing these objectives from expert knowledge must therefore be applied. In light of this, a study geared towards defining a procedure to build fuzzy sets and rules from traffic control experts' knowledge of a given city is being conducted by this research team. [ABSTRACT FROM AUTHOR]

We presented HITEC, an automated text classifier and its application categorize to English and German patent collections of WIPO under the IPC taxonomy. IPC covers all areas of technology and is currently used by the industrial property offices of many countries. Patent classification is indispensable for the retrieval of patent documents in the search for prior art. Such retrieval is crucial to patent-issuing authorities, potential inventors, research and development units, and others concerned with the application or development of technology. An efficient automated patent classifier is crucial component in providing an automated classification assistance system for categorizing patent applications in the IPC, that is a main aim at WIPO Fall et al., 2002. HITEC can be a prominent candidate for this purpose. [ABSTRACT FROM AUTHOR]

In this paper we have modeled a single trial ERP segment using the wavelet-based HMT model as an alternative to the averaging method for ERP analysis. The objective is to capture the dynamic changes of the ERP that the averaging method does not capture by allowing the estimation of the ERP for each single-trial. The HMT takes advantage of the persistence and clustering characteristics of the wavelet transform to model correlation structures within coefficients across scales, an approach that other wavelet-based methods do not use. These properties are suitable to model the statistical characteristics of an ERP signal. Due to the difficulty of knowing the signal distribution of an ERP segment, the use of a wavelet-based hidden Markov model is an advancement on the simple assumption of the Gaussian independent and identical distributions made by other noise reduction methods. This distribution is approximated using the Gaussian mixture included in the model. Experimental results with synthetic signals show that the HMT tree model will achieve a measurable improvement in the SNR of the estimate. When used in real ERP signals, the HMT produces estimates which preserve many of the features that are apparent in the noisy signal. In contrast, another wavelet-based method, the soft-thresholding method, appears to be inferior in preserving these features. [ABSTRACT FROM AUTHOR]

It was shown that recurrent neural networks with relatively simple architectures are capable of detecting seizure-like activity in multi-channel EEG records. Furthermore, the results obtained from the three different architectures studied were fairly consistent in locating the generators of seizure-like activity based on the patient's subdural EEG montage. If these results are verified by other studies, then these neural networks may serve as effective functional engines in more sophisticated seizure detection utilities. Additional work to increase the detection ability of these networks needs to be undertaken, especially with regard to the separation of the EEG data recorded at each electrode. Also, a means by which a utility using these neural networks can be made more adaptive to a variety of seizures from a wide patient population needs to be examined. [ABSTRACT FROM AUTHOR]

We have developed a new self-organizing network model of concept formation by using neurons with dynamic and spatial changing weights. The proposed network can recognize concepts using incomplete information of concepts at current degree of understanding and develop its synaptic connections of inner informational structure to discover new information of concepts and finally the complete information. [ABSTRACT FROM AUTHOR]