Artificial Neural Network Pattern Recognition Research Articles

Turbofan engine health status prediction with neural network pattern recognition and automated feature engineering

PurposeThis study aims to present the concept of aircraft turbofan engine health status prediction with artificial neural network (ANN) pattern recognition but augmented with automated features engineering (AFE).Design/methodology/approachThe main concept of engine health status prediction was based on three case studies and a validation process. The first two were performed on the engine health status parameters, namely, performance margin and specific fuel consumption margin. The third one was generated and created for the engine performance and safety data, specifically created for the final test. The final validation of the neural network pattern recognition was the validation of the proposed neural network architecture in comparison to the machine learning classification algorithms. All studies were conducted for ANN, which was a two-layer feedforward network architecture with pattern recognition. All case studies and tests were performed for both simple pattern recognition network and network augmented with automated feature engineering (AFE).FindingsThe greatest achievement of this elaboration is the presentation of how on the basis of the real-life engine operational data, the entire process of engine status prediction might be conducted with the application of the neural network pattern recognition process augmented with AFE.Practical implicationsThis research could be implemented into the engine maintenance strategy and planning. Engine health status prediction based on ANN augmented with AFE is an extremely strong tool in aircraft accident and incident prevention.Originality/valueAlthough turbofan engine health status prediction with ANN is not a novel approach, what is absolutely worth emphasizing is the fact that contrary to other publications this research was based on genuine, real engine performance operational data as well as AFE methodology, which makes the entire research very reliable. This is also the reason the prediction results reflect the effect of the real engine wear and deterioration process.

Fault Detection and Normal Operating Condition in Power Transformers via Pattern Recognition Artificial Neural Network

Aging, degradation, or damage to internal insulation materials often contribute to transformer failures. Furthermore, combustible gases can be produced when these insulation materials experience thermal or electrical stresses. This paper presents an artificial neural network for pattern recognition (PRN) to classify the operating conditions of power transformers (normal, thermal faults, and electrical faults) depending on the combustible gases present in them. Two network configurations were presented, one with five and the other with ten neurons in the hidden layer. The main advantage of applying this model through artificial neural networks is its ability to capture the nonlinear characteristics of the samples under study, thus avoiding the need for iterative procedures. The effectiveness and applicability of the proposed methodology were evaluated on 815 real data samples. Based on the results, the PRN performed well in both training and validation (for samples that were not part of the training), with a mean squared error (MSE) close to expected (0.001). The network was able to classify the samples with a 98% accuracy rate of the 815 samples presented and with 100% accuracy in validation, showing that the methodology developed is capable of acting as a tool for diagnosing the operability of power transformers.

Efficient water-related failure detection in PEM fuel cells: Combining a PEMFCs fractional order impedance model with FFT-PWM techniques and artificial neural network classification

Water management and early detection of faults in proton exchange membrane fuel cells (PEMFCs) are among the most critical constraints that limit the optimal spread of this type of energy. Consequently, it is necessary to enhance the reliability and durability of PEMFCs by developing an approach to diagnose and identify water failure modes. This paper proposes an effective and simple method to detect, diagnose, and classify various water failure modes in PEMFCs using a hybrid diagnostic approach. This approach combines the PEMFC fractional order impedance model (FOIM) with fast Fourier transform pulse width modulation (FFT-PWM) techniques and artificial neural network pattern recognition (ANN-PR) classification. The results show an accurate match between the electrochemical impedance spectroscopy (EIS) experimental data, the Nyquist impedance spectra of FOIM, and the FFT-PWM algorithm as a proposed alternative technique to EIS measurements. Learning of ANN-PR was performed using the frequency spectrum amplitude (FSA) database of the voltage and current signals produced by the PEMFCs FOIM DC/DC boost converter, which was generated using the FFT-PWM algorithm. The ANN-PR achieved low values for error accuracy, with the Low Square Error and Learning Error reaching 6.676 × 10−19 and 1.888 × 10−16, respectively. The elements inside the confusion matrix and the rest of the matrices confirm that the proposed model's accuracy, precision, recall, and high F1 score reached 100%. Furthermore, all predictions made by the ANN-PR model were consistently accurate across all areas of failure detection. Overall, the proposed method helps in analyzing, diagnosing, and classifying fuel cell failure modes such as flooding and drying, which may simplify the health assessment of PEMFC.

Pattern recognition describing spatio-temporal drivers of catchment classification for water quality

Classification using spatial data is foundational for hydrological modelling, particularly for ungauged areas. However, models developed from classified land use drivers deliver inconsistent water quality results for the same land uses and hinder decision-making guided by those models. This paper explores whether the temporal variation of water quality drivers, such as season and flow, influence inconsistency in the classification, and whether variability is captured in spatial datasets that include original vegetation to represent the variability of biotic responses in areas mapped with the same land use. An Artificial Neural Network Pattern Recognition (ANN-PR) method is used to match catchments by Dissolved Inorganic Nitrogen (DIN) patterns in water quality datasets partitioned into Wet vs Dry Seasons and Increasing vs Retreating flows. Explainable artificial intelligence approaches are then used to classify catchments via spatial feature datasets for each catchment. Catchments matched for sharing patterns in both spatial data and DIN datasets were corroborated and the benefit of partitioning the observed DIN dataset evaluated using Kruskal Wallis method. The highest corroboration rates for spatial data classification with DIN classification were achieved with seasonal partitioning of water quality datasets and significant independence (p < 0.001 to 0.026) from non-partitioned datasets was achieved. This study demonstrated that DIN patterns fall into three categories suited to classification under differing temporal scales with corresponding vegetation types as the indicators. Categories 1 and 3 included dominance of woodlands in their datasets and catchments suited to classify together change depending on temporal scale of the data. Category 2 catchments were dominated by vineforest and classified catchments did not change under different temporal scales. This demonstrates that including original vegetation as a proxy for differences in DIN patterns will help guide future classification where only spatially mapped data is available for ungauged catchments and will better inform data needs for water modelling.

Automated Stain‐Free Holographic Image‐Based Phenotypic Classification of Elliptical Cancer Cells

Image‐based stain‐free elliptical cancer cell classification is very challenging due to interclass morphological similarity. Herein, the classification of three types of cancer cell lines (lung, breast, and skin) by feature‐based machine learning and image‐based deep learning with a convolutional neural network (CNN) is addressed. Digital holography in a microscopic configuration is used to obtain stain‐free quantitative phase images representing the intracellular content and morphology of cells. In feature‐based classification, several features related to both the intracellular material and thickness of cancer cells are extracted, followed by the feature selection and the training of random forest, support vector machine, and pattern recognition artificial neural networks. For image‐based classification, two types of deep learning CNN models are trained: skip connections (Resnet) and without the skip connection. The accuracy of the two strategies is analyzed and the deep learning strategy outperforms feature‐based classification by about 9% with the 10‐fold cross‐validation evaluation.

A Novel Hyperspectral Method to Detect Moldy Core in Apple Fruits.

An innovative low-cost device based on hyperspectral spectroscopy in the near infrared (NIR) spectral region is proposed for the non-invasive detection of moldy core (MC) in apples. The system, based on light collection by an integrating sphere, was tested on 70 apples cultivar (cv) Golden Delicious infected by Alternaria alternata, one of the main pathogens responsible for MC disease. Apples were sampled in vertical and horizontal positions during five measurement rounds in 13 days’ time, and 700 spectral signatures were collected. Spectral correlation together with transmittance temporal patterns and ANOVA showed that the spectral region from 863.38 to 877.69 nm was most linked to MC presence. Then, two binary classification models based on Artificial Neural Network Pattern Recognition (ANN-AP) and Bagging Classifier (BC) with decision trees were developed, revealing a better detection capability by ANN-AP, especially in the early stage of infection, where the predictive accuracy was 100% at round 1 and 97.15% at round 2. In subsequent rounds, the classification results were similar in ANN-AP and BC models. The system proposed surpassed previous MC detection methods, needing only one measurement per fruit, while further research is needed to extend it to different cultivars or fruits.

Hybrid machine learning approaches for classification and detection of fractures in carbonate reservoir

Identifying reservoir fractures has been a challenge due to its significant influence in drilling and production, especially in highly complex carbonate reservoirs. In this paper, one of the well-known carbonate reservoirs, Asmari reservoir located in the Middle East, was considered and its fracture zones were classified as a machine learning nonlinear problem. First, a dataset was created which included conventional logs and fracture points from oil-based micro imager (OBMI) and oil mud reservoir imager (OMRI) with two circumferential and ultrasonic borehole imagers data from four wells. Next, the initial data were considered and the outliers were eliminated. The feature selection was carried out among conventional logs through pattern recognition artificial neural network (ANN) with the second version of the non-dominated sorting Genetic algorithm (NSGA-II). As a result of ANN-NSGA-II, six factors were selected and fed for classification methods. The targets of methods were three classes including no fracture, low fracture, and high fracture zones based on the user-defined thresholds. According to classification accuracies and confusion matrices, the support vector machine (SVM) with cubic kernel function outperformed the fine decision trees, quadratic discriminant analysis, and K-nearest neighbor (KNN) classifier methods. After determining the highest rank by SVM, five other kernel functions were tried as SVM functions. The radial basis function (RBF) with SVM showed a more reliable classification in comparison with the others. Regarding the objective of the present study, means introduced the most effective fracture classification. SVM was run with three neutral-inspired optimization producers including particle swarm optimization (PSO) and two newly presented grasshopper optimization algorithm (GOA) and grey wolf optimizer (GWO) methods. The final results indicated that (1) the hybrid SVM (RBF)-GWO offered superior accuracies in different population of algorithms, (2) classification zones with no fracture and low fracture (classes 1 and 2) had minimum misclassification through SVM-GWO while the high fracture points (class 3) revealed the highest accuracies with SVM-PSO. Also, SVM-GOA offered the best solution in several iterations and higher time, while, SVM-GWO and SVM-PSO seek SVM classification response faster and are thus recommended for fracture detection.

Rule-Based Anomaly Detection Model with Stateful Correlation Enhancing Mobile Network Security

The global Signalling System No. 7 (SS7) network protocol standard has been developed and regulated based only on trusted partner networks. The SS7 network protocol by design neither secures the communication channel nor verifies the entire network peers. The SS7 network protocol used in telecommunications has deficiencies that include verification of actual subscribers, precise location, subscriber’s belonging to a network, absence of illegitimate message filtering mechanism, and configuration deficiencies in home routing networks. Attackers can take advantage of these deficiencies and exploit them to impose threats such as subscriber or network data disclosure, intercept mobile traffic, perform account frauds, track subscriber location, and deny services. Existing methods are unable to identify suspicious hosts as they use a minimal number of network parameters. So, there is a vital need to overcome these deficiencies to detect the abnormal behaviour of users and hence mitigate security attacks in a mobile network. This research proposes a model for anomaly detection in mobile networks based on Rule-based filtering with stateful correlation. The performance of the proposed method is evaluated using synthetic datasets. Results show that the proposed anomaly detection model performs 0.37% better in terms of security attack detection rate, 24.25% better in terms of false alarm rate, and 31.45% better in terms of true positive rate when compared with the existing pattern recognition Artificial Neural Network (ANN) algorithm.

Rotating Machinery Fault Diagnosis Method Based on Improved Semisupervised Generative Confrontation Network

Error diagnosis and detection have become important in modern production due to the importance of spinning equipment. Artificial neural network pattern recognition methods are widely utilized in rotating equipment fault detection. These methods often need a large quantity of sample data to train the model; however, sample data (especially fault samples) are uncommon in engineering. Preliminary work focuses on dimensionality reduction for big data sets using semisupervised methods. The rotary machine’s polar coordinate signal is used to build a GAN network structure. ANN and tiny samples are utilized to identify DCGAN model flaws. The time-conditional generative adversarial network is proposed for one-dimensional vibration signal defect identification under data imbalance. Finally, auxiliary samples are gathered under similar conditions, and CCNs learn about target sample characteristics. Convolutional neural networks handle the problem of defect identification with small samples in different ways. In high-dimensional data sets with nonlinearities, low fault type recognition rates and fewer marked fault samples may be addressed using kernel semisupervised local Fisher discriminant analysis. The SELF method is used to build the optimum projection transformation matrix from the data set. The KNN classifier then learns low-dimensional features and detects an error kind. Because DCGAN training is unstable and the results are incorrect, an improved deep convolutional generative adversarial network (IDCGAN) is proposed. The tests indicate that the IDCGAN generates more real samples and solves the problem of defect identification in small samples. Time-conditional generation adversarial network data improvement lowers fault diagnosis effort and deep learning model complexity. The TCGAN and CNN are combined to provide superior fault detection under data imbalance. Modeling and experiments demonstrate TCGAN’s use and superiority.

Bionic arm: Mapping of elbow and wrist flexion using neural network and fuzzy logic

Cases on limb amputation necessitate the use of Transhumeral bionic for artificial limb rehabilitation, which is controlled using Electromyographic (EMG) signals from the muscles. Before the implementation of EMG control, a mapping between the movements of an arm to the angle formed at the corresponding joints is essential to be made. Most of the works in the field of Bionics use Supervised Machine Learning models, chiefly Classification, to map muscle flexion signals to joint actuations in the bionic arm. Ample literature is also there, which uses fuzzy logic for mapping. However, there are very few literatures that compare these two methods of mapping. In this article, 2 models have been discussed regarding the mapping, and their effectiveness is compared. The first model captures elbow and wrist flexion and maps them to their respective angular displacements of joints using a fuzzy logic model. In the second model, a Pattern Recognition Artificial Neural Network (ANN) model under Supervised Machine Learning is incorporated to map elbow and wrist flexion to the corresponding joint angular displacement. The ANN is trained with elbow and wrist joint flexion values and its corresponding joint angles data, optimized, and tested in real-time. This model is verified by comparing the joint angles of a test person (measured using Goniometers) with the joint angles of Bionic models made (using a 360° protractor sheet). The second model gave the insight that supervised machine learning models provide an accurate mapping to the joint flexion in the field of bionics.

Classification of catchments for nitrogen using Artificial Neural Network Pattern Recognition and spatial data

In hydrological modelling, classification of catchments is a fundamental task for overcoming deficits in observational datasets. Most attention on this issue has focussed on identifying the catchments with similar hydrological responses for streamflow. Yet, effective methods for catchment classification are currently lacking in respect to Dissolved Inorganic Nitrogen (DIN), a water quality constituent that, at increasing concentrations, is threatening nutrient sensitive environments. Pattern recognition, using standard Artificial Neural Network algorithm is applied, as a novel approach to classify datasets that are considered to be suitable proxies for biological and anthropogenic drivers of observed DIN releases. Eleven gauged Great Barrier Reef (GBR) catchments within Queensland Australia are classified using spatial datasets extracted from ecosystem (e.g. original ecosystem responses to biogeographic, land zone, land form, and soil type attributes) and land use maps. To evaluate the performance of the examined spatial datasets as a proxy for deductive classification, the classification process is repeated inductively, using observed DIN and streamflow data from gauging stations. The ANN-PR method is seen to generate the same classification score format for the differing dataset types, and this facilitates a direct comparison for model output for observed data corroborations. The Kruskal-Wallis test for independence, at p > 0.05, identifies the deductive classification approach as a predictor for classification using DIN observations, which lacks an independence from each other at a p value of 0.01 and 0.02. This study concludes that an ANN-PR method can integrate the ecosystem and land use mapping data to deductively classify the GBR catchments into four regions that also have similar patterns of DIN concentrations. Due to the uniform availability of the mapping data, the findings provide a sound basis for further investigations into the transposing of knowledge from gauged catchments to ungauged areas.

Artificial Neural Network Models for Coronary Artery Disease

Background: Coronary artery disease (CAD) is an important cause of mortality and morbidity globally. Objective : The early prediction of the CAD would be valuable in identifying individuals at risk, and in focusing resources on its prevention. In this paper, we aimed to establish a diagnostic model to predict CAD by using three approaches of ANN (pattern recognition-ANN, LVQ-ANN, and competitive ANN). Methods: One promising method for early prediction of disease based on risk factors is machine learning. Among different machine learning algorithms, the artificial neural network (ANN) algorithms have been applied widely in medicine and a variety of real-world classifications. ANN is a non-linear computational model that is inspired by the human brain to analyze and process complex datasets. Results: Different methods of ANN that are investigated in this paper indicate in both pattern recognition ANN and LVQ-ANN methods, the predictions of Angiography+ class have high accuracy. Moreover, in CNN, the correlations between the individuals in cluster ”c” with the class of Angiography+ are strongly high. This accuracy indicates the significant difference among some of the input features in Angiography+ class and the other two output classes. A comparison among the chosen weights in these three methods in separating control class and Angiography+ shows that hs-CRP, FSG, and WBC are the most substantial excitatory weights in recognizing the Angiography+ individuals although, HDL-C and MCH are determined as inhibitory weights. Furthermore, the effect of decomposition of a multi-class problem to a set of binary classes and random sampling on the accuracy of the diagnostic model is investigated. Conclusion: This study confirms that pattern recognition-ANN had the most accuracy of performance among different methods of ANN. This is due to the back-propagation procedure in which the network classifies input variables based on labeled classes. The results of binarization show that decomposition of the multi-class set to binary sets could achieve higher accuracy.

Classification of impedance cardiography dZ/dt complex subtypes using pattern recognition artificial neural networks.

In impedance cardiography (ICG), the detection of dZ/dt signal (ICG) characteristic points, especially the X point, is a crucial step for the calculation of hemodynamic parameters such as stroke volume (SV) and cardiac output (CO). Unfortunately, for beat-to-beat calculations, the accuracy of the detection is affected by the variability of the ICG complex subtypes. Thus, in this work, automated classification of ICG complexes is proposed to support the detection of ICG characteristic points and the extraction of hemodynamic parameters according to several existing subtypes. A novel pattern recognition artificial neural network (PRANN) approach was implemented, and a divide-and-conquer strategy was used to identify the five different waveforms of the ICG complex waveform with output nodes no greater than 3. The PRANN was trained, tested and validated using a dataset from four volunteers from a measurement of eight electrodes. Once the training was satisfactory, the deployed network was validated on two other datasets that were completely different from the training dataset. As an additional performance validation of the PRANN, each dataset included four volunteers for a total of eight volunteers. The results show an average accuracy of 96% in classifying ICG complex subtypes with only a decrease in the accuracy to 83 and 80% on the validation datasets. This work indicates that the PRANN is a promising method for automated classification of ICG subtypes, facilitating the investigation of the extraction of hemodynamic parameters from beat-to-beat dZ/dt complexes.

Implementation of coupled pattern recognition and regression artificial neural networks for mass estimation of headless‐shell‐on shrimp with random postures

AbstractIn an actual sorting process, shrimps are fed to a machine‐vision‐based sorter at random postures. This study proposed an Enhanced Artificial Neural Network (E‐ANN) coupled with a Pattern Recognition ANN (P‐ANN) model to overcome the posture‐specificity of the regression ANN model commonly used for mass estimation of the headless‐shell‐on (HSO) shrimps. Images of 103 shrimps with seven different postures were used. The similarity of any shrimp image to the reference shrimp postures (i.e., extended‐legs, collapsed‐legs, curl body, and dorsal body) was determined by the P‐ANN model and used as additional input, besides the area and perimeter. The coupled‐ANN model could accurately estimate the mass of shrimps with random postures (R2 = 0.70 to 0.88 and MRE = −2.62 to 2.97%) within ~10 ms per shrimp, which is practical to use in an automatic shrimp sorting system based on machine vision technique. Further enhancement of the model performance could be achieved by adding color and texture features to distinguish different shrimp parts (e.g., body, legs, and tail).Practical ApplicationsThis research shows the potential of using a pattern recognition artificial neural network (P‐ANN) to cut the limitation of the conventional regression ANN model based on area and perimeter for shrimp mass estimation, which is specific to shrimp posture (or shape). Since the coupled P‐ANN and enhanced artificial neural network (E‐ANN) used a relatively short processing time (~10 ms per shrimp) and yielded less than 1 g error in the mass estimation of the headless‐shell‐on (HSO) shrimps with random postures, it is possible to incorporate this coupled ANN model in an automatic shrimp sorting system based on machine vision technique.

A data-based hybrid model for complex fuel chemistry acceleration at high temperatures

During their high-temperature oxidation, complex hydrocarbons and their early fragments are short-lived and figure prominently only during the pyrolysis stage. However, they are quickly replaced by smaller hydrocarbons at the onset of the oxidation stage, resulting in simpler chemistry requirements past pyrolysis. In this study, we develop a data-based hybrid chemistry approach to accelerate chemistry integration for complex fuels. The approach is based on tracking the evolution of chemistry through representative species for the pyrolysis and coupling their reactions with simpler foundational chemistry. The selection of these representative species is implemented using principal component analysis (PCA) based on simulation data. The description of chemistry for the representative species is implemented using an artificial neural network (ANN) model for their reaction rates followed by the description of their chemistry using a foundational chemistry model. The selection of the transition between these models is trained a priori using an ANN pattern recognition classifier. This data-based hybrid chemistry acceleration model is demonstrated for three fuels: n-dodecane, n-heptane and n-decane and investigated with two foundational chemistry, C0–C2 and C0–C4, models. The hybrid scheme results in computational saving, up to one order of magnitude for n-dodecane, two orders of magnitudes for n-heptane, and three orders of magnitudes for n-decane. The accuracy and saving in computational cost depend on the number of selected species and the size of the used foundational chemistry. The hybrid model coupled with the more detailed C0–C4 foundational performs, overall, better than the one coupled with the C0–C2 foundational chemistry.

Trace Detection of Nerve Agent Simulant in the Fuel Vapour Environment using Metal Oxide Surface Acoustic Wave E Nose

Nerve agents are often used at the military warfront, where diesel is a very common interferant. In the present work, a group of surface acoustic wave (SAW) sensors, called E-Nose with dissimilar sensing layers is developed for the recognition of the mixture of diesel and dimethyl methylphosphonate (DMMP) vapors. The exposure of DMMP and diesel vapors is kept at ppb and ppm levels respectively. Varied response patterns of DMMP and diesel vapors were obtained by SAW E-nose. Principal component analysis (PCA) has been used to extract features from the response curves of SAW sensors. Artificial Neural Network pattern recognition has been implemented to identify the precise detection of DMMP vapors in the binary mixture of DMMP and diesel. The effect of pre-processing (using PCA) the raw data before feeding it to artificial neural network is also studied.

Simultaneous Flow and Thermal Conductivity Sensing on a Single Chip Using Artificial Neural Networks

A thin-film CMOS MEMS thermal sensor has been designed, fabricated and tested with the addition of through-membrane isolating holes. These holes have been shown to enhance the discrimination towards gases with differing thermal conductivity in the presence of flow. Using three on-membrane resistors as inputs, linear statistical methods alongside Artificial Neural Network pattern recognition techniques have been investigated for decoupling the two parameters of thermal conductivity and flow rate using a single sensor. In addition to this, the addition of the membrane holes increases the sensitivity towards flow rate by 10 times and the sensitivity towards thermal conductivity by 2 times. This sensor design coupled with well-known post-processing techniques will enable a new generation of multi-parameter sensing solutions.

Sensing Selectivity Enhancement of Palladium Oxide Toward VOCs Using Characteristic Response Features below 250oc

Breath analysis has received much attention for its non-invasive monitoring of human pathological conditions. The exhaled volatile organic compounds (VOCs) from patients can be utilized as potential biomarkers of specific diseases and disorders. For instance, acetone can be the indicator of diabetes and ethanol is an important interfering gas during the breath analysis. This work focuses on the improvement of the gas sensing selectivity for ethanol and acetone, by taking advantage of the intriguing redox behavior of palladium oxide (PdO) in the temperature range between 100 and 250oC. Metal oxide semiconductor (MOS) gas sensors have been extensively studied because of their rapid response and high sensitivity. Furthermore, fabrication of MOS gas sensors is simple and compatible to existing semiconductor fabrication processes. In this study, sputter-deposited PdO nanoflake thin films were used as the sensing material. PdO is a p-type semiconductor with a band gap of 2.2eV and thermally stable up to 800oC in atmosphere. In addition, PdO has a high catalytic reactivity toward reducing gases, such as CO and CH4. For the application of gas sensing, PdO is a good sensitizer in gas sensors for sensitivity and selectivity improvement. According to our previous works, PdO demonstrates characteristic sensing responses toward reducing gases due to its distinctive reduction and reoxidation behaviors. The characteristic sensing response may be ascribed to the reduction of PdO and subsequent reoxidation. In general, the gas sensing principle based on MOS gas sensors is described by two sensing mechanism models: ionosorption and oxygen vacancy models. The characteristic redox property of PdO can modify the gas sensing behavior predicted by the two models. Moreover, adsorption of intermediate adspecies on the PdO sensor in operation also affects the sensing response. Conventional MOS sensors use the sensor signal, which is typically determined under a steady sensing condition, as the input variable for the data processing of gas classification. In this case, simple mathematical models, such as principle component analysis (PCA) and linear discriminant analysis (LDA), can be used to classify the input signals. However, these simple models have poor performance in selectivity improvement and thus have limited applications for MOS gas sensors. Since the PdO gas sensor demonstrates dynamic characteristic responses toward VOC gases, we extracted characteristic response features as the input variables for the artificial neural network (ANN) method to improve the selectivity of VOC gas sensing. Figure (a) shows the electrical sensing response of PdO exposed to a dry air gas mixture of VOCs at 250oC. Ammonia is also included in the study because it is a strong reducing agent and an important indicator of renal diseases and asthma. At the early stage of sensing, the sensing current promptly drops upon the exposure to the gases. The sensing current then gradually rises and reaches the equilibrium. The first prompt current drop can be ascribed to reduction of preadsorbed oxygen anions and PdO by VOCs on the basis of ionosorption and oxygen vacancy models. Reducing gases can react with oxygen anions and lattice oxygen atoms, releasing negative charges to the PdO sensor and thus decreasing its conductance. The latter current rise is a result of the formation of metal Pd nano-domain, which is a surface product of the PdO reduction. Chemical characterizations indicate that ethanol, acetone and ammonia undergo dehydrogenation on the PdO surface. These reducing gases can provide a plentiful amount of hydrogen to reduce the PdO substrate, depending on their bond strength and adsorption structure on the sensor. The reduction rate of the PdO substrate greatly depends on the surface coverage of hydrogen adatoms and the sensing temperature. As a result, the PdO sensor yields distinct response profiles when it is exposed to the studied gases at different temperatures. Two characteristic features extracted from the response profiles for the ANN method are depicted in figure (b). The two features are the slope of the conductance rise trace and the ratio of conductance change at two time points. The classification result is listed in table 1, demonstrating a high classification rate. In summary, the gas sensing behavior of PdO toward VOCs was studied. Characteristic features from the electrical response profiles were used as variables for the ANN pattern recognition algorithm to enhance the sensing selectivity of PdO toward VOCs. Figure 1

Distinguishing Petroleum (Crude Oil and Fuel) From Smoke Exposure within Populations Based on the Relative Blood Levels of Benzene, Toluene, Ethylbenzene, and Xylenes (BTEX), Styrene and 2,5-Dimethylfuran by Pattern Recognition Using Artificial Neural Networks.

Studies of exposure to petroleum (crude oil/fuel) often involve monitoring benzene, toluene, ethylbenzene, xylenes (BTEX), and styrene (BTEXS) because of their toxicity and gas-phase prevalence, where exposure is typically by inhalation. However, BTEXS levels in the general U.S. population are primarily from exposure to tobacco smoke, where smokers have blood levels on average up to eight times higher than nonsmokers. This work describes a method using partition theory and artificial neural network (ANN) pattern recognition to classify exposure source based on relative BTEXS and 2,5-dimethylfuran blood levels. A method using surrogate signatures to train the ANN was validated by comparing blood levels among cigarette smokers from the National Health and Nutrition Examination Survey (NHANES) with BTEXS and 2,5-dimethylfuran signatures derived from the smoke of machine-smoked cigarettes. Classification agreement for an ANN model trained with relative VOC levels was up to 99.8% for nonsmokers and 100.0% for smokers. As such, because there is limited blood level data on individuals exposed to crude oil/fuel, only surrogate signatures derived from crude oil and fuel were used for training the ANN. For the 2007-2008 NHANES data, the ANN model assigned 7 out of 1998 specimens (0.35%) and for the 2013-2014 NHANES data 12 out of 2906 specimens (0.41%) to the crude oil/fuel signature category.

Whose Balance Sheet Is This? Neural Networks for Bankss Pattern Recognition

The balance sheet is a snapshot that portraits the financial position of a firm at a specific point of time. Under the reasonable assumption that the financial position of a firm is unique and representative, we use a basic artificial neural network pattern recognition method on Colombian banks’ 2000-2014 monthly 25-account balance sheet data to test whether it is possible to classify them with fair accuracy. Results demonstrate that the chosen method is able to classify out-of-sample banks by learning the main features of their balance sheets, and with great accuracy. Results confirm that balance sheets are unique and representative for each bank, and that an artificial neural network is capable of recognizing a bank by its financial accounts. Further developments fostered by our findings may contribute to enhancing financial authorities’ supervision and oversight duties, especially in designing early-warning systems.