Improving Structure Damage Identification by Using ICA-ANN Based Sensitivity Analysis

Structural damage detection is very important for identifying and diagnosing the nature of the damage in an early stage so as to reduce catastrophic failures and prolong the service life of structures. In this paper, a novel approach is presented that integrates independent component analysis (ICA) and support vector machine (SVM). The procedure involves extracting independent components from measured sensor data through ICA and then using these signals as input data for a SVM classifier. The experiment presented employs the benchmark data from the University of British Columbia to examine the effectiveness of the method. Results showed that the accuracy of damage detection using the proposed method is significantly better than the approach by integrating ICA and ANN. Furthermore, the prediction output can be used to identify different types and levels of structure damages.

This study presents a combined artificial neural network (ANN) and multivariate approach for performance evaluation and optimization of gas refineries. This study introduces standard financial and non-financial indicators for performance evaluation of the gas refineries. Data are collected from gas balance sheets and the detailed statistics of gas refineries. Two cases have been considered for performance evaluation. In the first case the financial indicators and in the second case the financial and non-financial indicators are used and tested over five years period. The refineries are evaluated by data envelopment analysis (DEA), principal component analysis (PCA), numerical taxonomy and artificial neural network (ANN). Finally, a complete sensitivity analysis is performed for each stated method. The results show that DEA is more resistant to noise than other methods. Also, there is slight difference between results of financial and combined financial and operational indicators. This suggests the use of combined financial and operational indicators for future practical studies in gas refineries. This is the first study that presents an integrated approach for combined performance of financial and operational indicators in gas refineries.

Hybrid pattern recognition was put forward to discriminate paddy seeds of four different storage periods based on visible/near infrared reflectance spectroscopy (Vis/NIRS). The hybrid pattern recognition included extracting feature and building classifier. A total of 210 samples of paddy seeds, which belonged to four classes, were used for collecting Vis/NIR spectra (325- 1075 nm) using a field spectroradiometer. The hybrid pattern recognition was integrated with wavelet transform (WT), principal component analysis (PCA) and artificial neural networks (ANN) models. WT was used to eliminate noises and extract characteristic information from spectral data. The characteristic information could be visualized in principal components (PCs) space, in which the structures correlative with the storage periods could be discovered. The first eight PCs, which accounted for 99.94% of the raw spectral data variance, were used as input of the ANN mode, and the model yielded high discrimination accuracy rates of 100%, 100%, 100% and 90% for four classes' samples respectively.

SummaryFour different honey types (Acacia, Robinia pseudoacacia; chestnut, Castanea sativa; lime, Tilia sp.; sage, Salvia officinalis) from Croatia were analysed using an electronic nose to capture the aroma profile. Principal Component Analysis (PCA) and Artificial Neural Network (ANN) models were used to determine the botanical origin of honey samples. Overall accurate classification was over 90% which suggests that an electronic nose could be used as fast screening method for the determination of honey botanical origin.

Multitemporal Landsat TM imagery were classified to extract land cover information using principal component analysis (PCA) and backpropagation (BP) algorithm of artificial neural network. Data used are two Landsat TM data of in Jan. 1, 1991 (Data I) and May 9, 1994 (Data II). Twelve bands data were compressed to 4 bands data by the first and second PCA. Approximately 95 percent of the total variance of each Landsat TM data was included resulting from the first and second component analysis. Analyzed data through the PCA were classified by the BP training algorithm of artificial neural network. As a result of classification, it is concluded that this approach will become an attractive and effective method in extracting land cover or land use information using multitemporal Landsat TM data.

Artificial neural network model for predicting the local compression capacity of stirrups-confined concrete

Five models of a typical bridge pier at 0.45 scale were constructed in their original (1950's) as-built condition. Four were then retrofitted following several alternative schemes. In the structural laboratory at the University of British Columbia, each bent was subjected to sequences of lateral slow cyclic loading, which was increased for each sequence up to a high damage level. Between the sequences, corresponding to increasing levels of structural damage, forced vibration tests were performed on the bents with an instrumented hammer. Frequency domain analyses of the experimental data were performed to identify dynamic characteristics of each bent at each level of structural damage. The decreasing fundamental frequencies were related to the increasing ductility demands, the latter reflecting the structural damage of the bents. Sensitivity analyses were also performed to relate damping ratio changes to ductility levels imposed on the specimens.

The paper investigates whether the sensitivity analysis can be used not only as a tool to read the knowledge embedded in artificial neural networks (ANNs), but also as a tool to evaluate the effectiveness of ANN learning. The simulation of the Black-Scholes formula is employed for this object. The Black-Scholes formula, in which the mapping between the call price and five relevant variables is a mathematically closed form, is suitable for verifying the validity of the methodology of sensitivity analysis in reading ANN knowledge. As for the validity of evaluating the effectiveness of ANN learning, two different ANNs are set up, and their sensitivity analyses on learning patterns are compared. The experimental results show that both values of sensitivity analysis of ANNs and partial derivative of the Black-Scholes formula are consistent. Furthermore, they indicate that the sensitivity analysis can be used as a tool to evaluate the effectiveness of ANN learning.

Despite its significance for performance-based design, concrete’s tensile strength, which governs crack formation, structural damage, and the longevity of the structure, has usually received inadequate attention in predictive modeling compared to compressive strength. This paper proposes a complete machine learning approach to forecast the tensile strength of nano-engineered and sustainable concretes utilizing a novel database containing 500 data points and ten diverse input factors such as water-to-cement (w/c) ratio, curing time, coarse and fine aggregates, and content of other additives (basalt fiber, carbon nanotube, geopolymer binder, superplasticizer). The study employed and compared four state-of-the-art machine learning algorithms, including support vector regression (SVR), artificial neural networks (ANN), extreme gradient boosting (XGBoost), and a novel hybrid ensemble model (HEM) that uses a weighted meta-regressor to combine outputs from multiple base learners. The HEM model consistently outperformed the others in hold-out cross-validation, K-fold cross-validation, and generalization tests. It achieved the best predictive accuracy with a K-fold cross-validation composite score of 96. ANN came in second place, followed by SVR and XGBoost, with K-fold cross-validation ranking scores of 70, 53, and 25, respectively. In the generalization tests, HEM demonstrably followed the nonlinear trends in the tensile strength for w/c ratios (peak at 0.4 with 36.95 MPa), curing time (peak at 56 days with 38.49 MPa), and nano-clay content (peak at 5% with 38.86 MPa), surpassing the other models in every case. Sensitivity analysis using the HEM model revealed optimal input values for maximizing tensile strength, including cement content (378.7 kg/m2), w/c ratio (0.406), fine aggregate (830.5 kg/m2), coarse aggregate (1034.5 kg/m2), nano-clay (4.98%), carbon nanotubes (1.00%), geopolymer binder (18.18%), basalt fiber (2.56%), and superplasticizer (< 0.93%). The findings of this study offer not only a reasonable and precise model for forecasting tensile strength but also a predictive analytical framework for enhancing eco-friendly concrete mix designs. By combining sophisticated machine learning algorithms with materials science, the article contributes to the development of innovative, performance-oriented, more durable, and environmentally friendly infrastructure construction and maintenance.

The application of the TERT-IV system prototype by Tianjin University to identify two-phase flow regimes has been introduced. For the several typical flow regimes, the method of principal component analysis and artificial neural network to identify the two-phase flow regimes is presented, and that is proved to have higher recognition rate by experimental test. The research results show that the method is feasible using feature extraction and analysis data to identify two-phase flow regimes under different flow conditions, and prove that it is possible that on line monitor the transportation process of air/water two-phase flow using electrical resistance tomography (ERT) system.

A deep geological repository for radioactive waste, such as Andra's Cigéo project, requires long-term (persistent) monitoring. To achieve this goal, data from a network of sensors are acquired. This network is subject to deterioration over time due to environmental effects (radioactivity, mechanical deterioration of the cell, etc.), and it is paramount to assess each sensor's integrity and ensure data consistency to enable the precise monitoring of the facilities. Graph neural networks (GNNs) are suitable for detecting faulty sensors in complex networks because they accurately depict physical phenomena that occur in a system and take the sensor network's local structure into consideration in the predictions. In this work, we leveraged the availability of the experimental data acquired in Andra's Underground Research Laboratory (URL) to train a graph neural network for the assessment of data integrity. The experiment considered in this work emulated the thermal loading of a high-level waste (HLW) demonstrator cell (i.e., the heating of the containment cell by nuclear waste). Using real experiment data acquired in Andra's URL in a deep geological layer was one of the novelties of this work. The used model was a GNN that inputted the temperature field from the sensors (at the current and past steps) and returned the state of each individual sensor, i.e., faulty or not. The other novelty of this work lay in the application of the GraphSAGE model which was modified with elements of the Graph Net framework to detect faulty sensors, with up to half of the sensors in the network being faulty at once. This proportion of faulty sensors was explained by the use of distributed sensors (optic fiber) and the environmental effects on the cell. The GNNs trained on the experimental data were ultimately compared against other standard classification methods (thresholding, artificial neural networks, etc.), which demonstrated their effectiveness in the assessment of data integrity.

This paper presents a sensitivity and reliability analysis of a self-anchored suspension bridge by applying a new hybrid method proposed by the authors based on integration of the Latin hypercube sampling technique (LHS), artificial neural network (ANN), first-order reliability method (FORM), Pearson's linear correlation coefficient (PLCC), and Monte Carlo simulation with important sampling technique (MCS-IS). The framework consists of three stages of analysis: (1) selection of training, validation, and test datasets for establishing an ANN model by the LHS technique; (2) formulation of a performance function from the well-trained ANN model; and (3) sensitivity analysis using PLCC, identification of the most probabilistic failure point based on FORM, and estimation of the failure probability using the MCS-IS technique. Upon demonstration of its efficiency through analysis of a 12-story frame structure, the method is applied to sensitivity and reliability analysis of the Jiangxinzhou Bridge, a five-span self-anchored suspension bridge, in which both structural parameters and external loads are considered as random variables. The analysis identified a number of structural parameters, as well as external loads, that have a significant influence on structural serviceability and safety.

This paper presents a model for calculation of torsion capacity of the reinforced concrete beams using the artificial neural network. Considering the complex reaction of reinforced concrete beams under torsion moments, torsion strength of these beams is depended on different parameters; therefore using the artificial neural network is a proper method for estimating the torsion capacity of the beams. In the presented model the beam's dimensions, concrete compressive strength and longitudinal and traverse bars properties are the input data, and torsion capacity of the reinforced concrete beam is the output of the model. Also considering the neural network results, a sensitivity analysis is performed on the network layers weight, and the effect of different parameters is evaluated on the torsion strength of the reinforced concrete beams. According to the sensitivity analysis, properties of traverse steel have the most effect on torsion capacity of the beams.

The use of machine learning has grown in popularity in various disciplines. Despite the popularity, the apparent ‘black box’ nature of such tools continues to be an area of concern. In this article, we attempt to unravel the complexity of this black box by exploring the use of artificial neural networks (ANNs), coupled with graph theory, to model and interpret the spatial distribution of building damage from extreme wind events at a community level. Structural wind damage is a topic that is mostly well understood for how wind pressure translates to extreme loading on a structure, how debris can affect that loading and how specific social characteristics contribute to the overall population vulnerability. While these themes are widely accepted, they have proven difficult to model in a cohesive manner, which has led primarily to physical damage models considering wind loading only as it relates to structural capacity. We take advantage of this modelling difficulty to reflect on two different ANN models for predicting the spatial distribution of structural damage due to wind loading. Through graph theory analysis, we study the internal patterns of the apparent black box of artificial intelligence of the models and show that social parameters are key to predict structural damage.

Artificial neural networks (ANNs) excel at analyzing challenging data sets and can be exceptional tools for decision support in clinical environments. The present study pilots the use of ANNs for determining prognosis in neuro-oncology patients. To determine whether ANNs perform better at predicting 1-year survival in a group of patients with brain metastasis compared with traditional predictive tools. : ANNs were trained on a multi-institutional data set of radiosurgery patients to predict 1-year survival on the basis of several input factors. A single ANN, an ensemble of 5 ANNs, and logistic regression analyses were compared for efficacy. Sensitivity analysis was used to identify important variables in the ANN model. A total of 196 patients were divided up into training, testing, and validation data sets consisting of 98, 49, and 49 patients, respectively. Patients surviving at 1 year tended to be female (P = .001) and of good performance status (P = .01) and to have favorable primary tumor histology (P = .001). The pooled voting of 5 ANNs performed significantly better than the multivariate logistic regression model (P = .02), with areas under the curve of 84% and 75%, respectively. The ensemble also significantly outperformed 2 commonly used prognostic indexes. Primary tumor subtype and performance status were identified on sensitivity analysis to be the most important variables for the ANN. ANNs outperform traditional statistical tools and scoring indexes for predicting individual patient prognosis. Their facile implementation, robustness in the presence of missing data, and ability to continuously learn make them excellent choices for use in complicated clinical environments.