Unknown Test Samples Research Articles

Dynamic model-based intelligent fault diagnosis method for fault detection in a rod fastening rotor

A complete fault sample database is of great significance for the intelligent fault diagnosis method of rod fastening rotor. However, the lack of fault samples makes the fault diagnosis results unbelievable. To solve this issue, the dynamic model-based intelligent fault diagnosis method is established for a rod fastening rotor, and the fault sample database is enriched by numerical simulations. First, the lumped parameter model of the rod fastening rotor system is constructed and further updated using Euclidean Distance between measurement and numerical simulation of the intact system. Second, mathematical models of various fault types are incorporate into the updated model to obtain numerical simulation fault samples. Thirdly, the utilization of numerical simulation fault samples is severed as training data to the artificial intelligence (AI) models and the unknown measurement test samples will be finally classified. In this paper, Support Vector Machine, Random Forest, Bayesian Network and Decision Tree are selected as the typical AI models. Subsequently, the feasibility of classification is validated by the test bench of the rod fastening rotor system, and the problem of insufficient fault samples can be solved.

Simplified and Detailed Evaluations of the Uncertainty of the Measurement of Microbiological Contamination of Pharmaceutical Products.

The control of the microbial contamination of pharmaceutical products (PP) is crucial to ensure their safety and efficacy. The validity of the monitoring of such contamination depends on the uncertainty of this quantification. Highly uncertain quantifications due to the variability of determinations or the magnitude of systematic effects affecting microbial growth or other analytical operations make analysis unfit for the intended use. The quantification of the measurement uncertainty expressing the combined effects of all random and systematic effects affecting the analysis allows for a sound decision about quantification adequacy for their intended use. The complexity of the quantification of microbial analysis uncertainty led to the development of simplified ways of performing this evaluation. This work assesses the adequacy of the simplified quantification of the uncertainty of the determination of the microbial contamination of PP by log transforming microbial count and dilution factor of the test sample whose uncertainty is combined in a log scale using the uncertainty propagation law. This assessment is performed by a parallel novel bottom-up and accurate evaluation of microbial analysis uncertainty involving the Monte Carlo method simulation of the Poisson log-normal distribution of counts and of the normally distributed measured volumes involved in the analysis. Systematic effects are assessed and corrected on results to compensate for their impact on the determinations. Poisson regression is used to predict precision affecting determinations on unknown test samples. Simplified and detailed models of the uncertainty of the measurement of the microbial contamination of PP are provided, allowing objective comparisons of several determinations and those with a maximum contamination level. This work concludes that triplicate determinations are required to produce results with adequately low uncertainty and that simplified uncertainty quantification underevaluates or overevaluates the uncertainty from determinations based on low or high colony numbers, respectively. Therefore, detailed uncertainty evaluations are advised for determinations between 50 and 200% of PP's maximum admissible contamination value. User-friendly tools for detailed and simplified evaluations of the uncertainty of the measurement of microbial contamination of PP are provided together with the understanding of when simplifications are adequate.

Assessment of Beeswax Adulteration by Paraffin and Stearic Acid Using ATR-IR Spectroscopy and Multivariate Statistics-An Analytical Method to Detect Fraud.

A spectroscopic investigation of beeswax adulteration by paraffin and/or stearic acid was undertaken via Attenuated Total Reflectance Infra-Red spectroscopy (ATR-IR) combined with multivariate statistical analyses. Principal Component Analysis (PCA) was successfully applied for the first time as an exploratory tool for the differentiation among pure beeswax and adulterated beeswax by paraffin and stearic acid with detection limits (LOD) of ~5% and 1%, respectively. Partial Least Square (PLS) modelling was used to build chemometric models based on beeswax/paraffin and beeswax/stearic acid calibration mixtures and subsequently used to predict concentrations of paraffin and stearic acid on a set of unknown test samples. PLS predictions demonstrated that beeswax adulteration by paraffin is much more prominent (74%) than the one by stearic acid (26%) and that commercial beeswax products (candles, pearls, blocks, etc.) are more prone to adulteration (27%) than honeycomb-type samples (12.5%).

Contrastive Open Set Recognition

In conventional recognition tasks, models are only trained to recognize learned targets, but it is usually difficult to collect training examples of all potential categories. In the testing phase, when models receive test samples from unknown classes, they mistakenly classify the samples into known classes. Open set recognition (OSR) is a more realistic recognition task, which requires the classifier to detect unknown test samples while keeping a high classification accuracy of known classes. In this paper, we study how to improve the OSR performance of deep neural networks from the perspective of representation learning. We employ supervised contrastive learning to improve the quality of feature representations, propose a new supervised contrastive learning method that enables the model to learn from soft training targets, and design an OSR framework on its basis. With the proposed method, we are able to make use of label smoothing and mixup when training deep neural networks contrastively, so as to improve both the robustness of outlier detection in OSR tasks and the accuracy in conventional classification tasks. We validate our method on multiple benchmark datasets and testing scenarios, achieving experimental results that verify the effectiveness of the proposed method.

ID2S4FH: A Novel Framework of Intelligent Decision Support System for Fire Hazards

Modern societies and industrial sectors are serviced through storage and distribution centres (SDCs) such as supermarkets, malls, warehouses, etc. Large quantities of supplies are stocked here, e.g., food grains, clothes, shoes, pharmaceuticals, electronics, plastics, edible oils, electrical wires/equipment, petroleum products, painting materials, etc. Fires due to the burning of these materials are categorized into six classes, viz., Class A, Class B, Class C, Class D, Class K, and Class F. A fire is extinguished better when the right type of fire retardant is used. A thumb rule on firefighting also says, “never fight a fire if you do not know what is burning”. In this paper, we have proposed an Intelligent Decision Support System (ID2S4FH) to generate a real-time ‘fire-map’ of such SDCs during a fire hazard. We have interfaced six tin-oxide-based gas sensor elements, a temperature and humidity sensor, and a particulate matter (PM) sensor with microcontrollers to capture the real-time signature patterns of the ambient air. We burned sixteen different types of materials belonging to six classes of fire and created a dataset consisting of 2400 samples. The sensor array responses were then pre-processed and analysed using various classifiers trained in different analysis space domains. Among the classifiers, four classifiers achieved ‘all correct’ identification of the fire classes of 80 unknown test samples, and the lowest mean squared error (MSE) achieved was 2.81 × 10−3. During a fire hazard, our proposed ID2S4FH can generate real-time fire maps of SDCs and help firefighters to extinguish the fire using the appropriate fire retardant.

An IoT-Enabled E-Nose for Remote Detection and Monitoring of Airborne Pollution Hazards Using LoRa Network Protocol.

Detection and monitoring of airborne hazards using e-noses has been lifesaving and prevented accidents in real-world scenarios. E-noses generate unique signature patterns for various volatile organic compounds (VOCs) and, by leveraging artificial intelligence, detect the presence of various VOCs, gases, and smokes onsite. Widespread monitoring of airborne hazards across many remote locations is possible by creating a network of gas sensors using Internet connectivity, which consumes significant power. Long-range (LoRa)-based wireless networks do not require Internet connectivity while operating independently. Therefore, we propose a networked intelligent gas sensor system (N-IGSS) which uses a LoRa low-power wide-area networking protocol for real-time airborne pollution hazard detection and monitoring. We developed a gas sensor node by using an array of seven cross-selective tin-oxide-based metal-oxide semiconductor (MOX) gas sensor elements interfaced with a low-power microcontroller and a LoRa module. Experimentally, we exposed the sensor node to six classes i.e., five VOCs plus ambient air and as released by burning samples of tobacco, paints, carpets, alcohol, and incense sticks. Using the proposed two-stage analysis space transformation approach, the captured dataset was first preprocessed using the standardized linear discriminant analysis (SLDA) method. Four different classifiers, namely AdaBoost, XGBoost, Random Forest (RF), and Multi-Layer Perceptron (MLP), were then trained and tested in the SLDA transformation space. The proposed N-IGSS achieved "all correct" identification of 30 unknown test samples with a low mean squared error (MSE) of 1.42 × 10-4 over a distance of 590 m.

Spatial Upscaling-Based Algorithm for Detection and Estimation of Hazardous Gases

Recently, society/industry is getting smarter and sustainable through artificial intelligence-based solutions. However, this rapid advancement is also polluting our air ambience. Hence real-time detection and estimation of hazardous gases/odors in the air ambiance has become a critical need. In this paper, a convolutional neural network (CNN) based multi-element gas sensor arrays signature response analysis approach has been presented to achieve higher accuracy in detection and estimation of hazardous gases. Accordingly, the real-time gas sensor array responses are spatially upscaled and processed on the edge, using lightweight CNNs. For the verification of our hypothesis, we have used a four-element metal-oxide semi-conductor (MOS)-based thick-film gas sensor array, fabricated by our group, by using SnO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> , ZnO, MoO, CdS materials for detection and estimation of four target hazardous gases, viz., acetone, carbon-tetrachloride, ethyl-methyl-ketone, and xylene. The four-element (2×2) raw sensor responses are first upscaled to 6×6 responses and a lightweight CNN is trained on 42 samples of 6×6 input vectors. The trained system is then tested using 16 unknown (not used during training) test samples of the considered gases/odors. All the 16 test samples are detected correctly. The Mean Squared Error (MSEs) of detection has been 1.42×10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-14</sup> while the estimation accuracy of 2.43×10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-3</sup> were achieved for the considered gases. Our designed system is generic in design and can be extended to other gases/odors of interest.

Accurate detection of porosity in glass fiber reinforced polymers by terahertz spectroscopy

Pores are inevitably produced during the production of glass fiber reinforced polymers (GFRP). Quantitative characterization of porosity is a most important aspect of performance evaluation of GFRP. Herein, we present a novel strategy for porosity detection of GFRP based on the interaction mechanism between terahertz (THz) wave and porous GFRP (porosity: 0.29%–4.01%; pore size: about 20–600 μm). By using the transmission and absorption spectra of GFRP, a porosity prediction model is established by combining the supervised learning approach of support vector regression (SVR) and ensemble methods, which can predict the porosity of unknown test samples with a coefficient of determination R2 = 0.976 and root mean square error RMSE = 0.174%. Conversely, THz transmission and absorption spectra for porous GFRP are successfully reconstructed by SVR and the ensemble methods. The results indicate that THz spectroscopy in combination with SVR and the ensemble methods is robust and accurate in porosity analysis and could play a significant role in industrial applications where nondestructive on-line detection of porosity in polymer composites is desirable.

Abductive statistical methods improve the results of calibration curve bioassays: An example of determining zinc bioavailability in broiler chickens.

In this paper, we discuss the theory behind calibration curve experiments and their application to a zinc (Zn) bioavailability study with broiler chickens. Seven replicates of 16 male commercial broiler chicks were fed starter diets for 14 days. Six diets had different levels of a potential Zn source and one was a positive control with standard industry levels of Zn for comparison. Four commonly used methods of calculating bioavailability means and confidence intervals (CI) from a calibration curve (standard curve) experiment to estimate the bioavailability of a new zinc source in broiler chickens were compared. The methods compared were the following: 1) the Counter-Intuitive Method uses a multiple-range test to compare unknown test and standard samples; 2) the Intuitive Method uses standard linear regression and inverts the equation to predict Zn bioavailability for each replicate of test samples; 3) the Abductive Method uses Graybill's Equation, based on theory and observation, to estimate CI's; and 4) the Sophistic Method uses reverse regression, and calculates Zn bioavailability values directly from the equation. The Counter-Intuitive Method only gives information about which standards the test samples are, or are not, significantly different from respectively (average available Zn not predicted). The Intuitive Method ignores error about the standard curve and theoretically cannot estimate the CI directly ( = 107.5 ± 15.8 mg Zn/kg). The Sophistic Method underestimates and overestimates the test sample mean values above and below the mean of the standards, respectively ( = 96.6 mg Zn/kg). The Abductive Method has an advantage over the other methods: The mean prediction estimation is consistent with theory (107.5 ± 6.1 mg Zn/kg; ). When test or “unknown” samples are near the mean of the standard samples, the CI is smaller than when near the extremes of the calibration curve. When calibration curve error is small (R2 > approximately 0.95), there is little advantage to using the Abductive Method, but when calibration curve error is larger, as in many bioassays with growing animals, the Abductive Method improves the accuracy of the CI calculations. The Abductive Method was used to demonstrate the influence of the number of replicate samples on experimental power and cost.

Recognition of gasoline in fire debris using machine learning: Part II, application of a neural network

The recognition of ignitable liquid (IL) residues in fire debris is a resource intensive but key part of an arson investigation. Due to the highly diverse and heavily loaded chemical matrix of fire debris samples, combined with the broad chemical composition of IL, the interpretation of the laboratory analysis results is a very challenging task for the forensic examiner. Fire debris samples are commonly analyzed using gas chromatography coupled to mass spectrometry (GC-MS). This method delivers both the total ion chromatogram (TIC) with the individually separated compounds and the underlying mass spectrum of each of the separated compounds. In this study, a completely new approach for the recognition of gasoline in fire debris samples is presented. First, the GC-MS data, including retention time, signal intensity, and mass spectrum is converted into a bitmap image. Five different data-to-image conversion approaches are tested, and their advantages and limitations are discussed. Subsequently, a convolutional neural network (CNN) is utilized to allocate the generated images to the classes “with gasoline” or “without gasoline”. The applied approaches to generate a digital image and the pattern recognition of the CNN perform very well in the classification of unknown test samples. Depending on the data-to-image generation approach used, the rate of correct sample classification in the test dataset is between 95% and 98%. The machine learning approach in this study, as well as the complementary method presented in an accompanying article, are not only useful for the recognition of gasoline in fire debris but are equally applicable to any additional areas in which the interpretation of complex chromatographic and mass spectrometric is required.

Machine learning-assisted array from fluorescent conjugated microporous polymers for multiple explosives recognition

The fluorescent properties of conjugated microporous polyphenylene (CMPs) were tuned through a wide range by inclusion of small amount of comonomer as chromophore in the network. The multi-color CMPs were used for explosives sensing and demonstrated broad sensitivity (ranging from −0.01888 μM−1 to −0.00467 μM−1) and LODs (ranging from 31.0 nM to 125.3 nM) against thirteen explosive compounds including nitroaromatics (NACs), nitramines (NAMs) and nitrogen-rich heterocycles (NRHCs). The CMPs were also developed as a sensor array for discrimination of thirteen explosives, specifically including NT, p-DNB, DNT, TNT, TNP, TNR, RDX, HMX, CL-20, FOX-7, NTO, DABT and DHT. By using classical statistical method “Linear Discriminant Analysis (LDA)”, the thirteen explosives at a fixed concentration were completely discriminated and unknown test samples were indentied with 88% classification accuracy. Moreover, explosives in different concentrations and the mixtures of explosives were also successfully classified. Compared with LDA, Machine Learning algorithms have significant advantages in analyzing the array-based sensing data. Different Machine Learning models for pattern recognition have also been implemented and discussed here and much higher accuracy (96% for “neural network”) can be achieved in predicting unknown test samples after training.

A review on state-of-the-art Automatic Speaker verification system from spoofing and anti-spoofing perspective

Background/Objectives: The anti-spoofing measures are blooming with an aim to protect the Automatic Speaker Verification systems from susceptible spoofing attacks. This review is an amalgam of the possible attack types, the datasets required, the renowned feature representation techniques, modeling algorithms involving machine learning, and score normalization techniques. Method/Findings: A detailed analysis of existing datasets is carried based on the total speaker samples, the number of speakers, and source of availability- open or licensed. This may foster choosing the right dataset for building the anti-spoofing frameworks. Further, the feature extraction schemes are elaborated with an intention to cover the vast span of features existing in various parts of raw speech for obtaining speaker-specific traits. Further, the machine learning algorithms ranging from discriminative to generative to mixed form are explored for seeking the right algorithm in specific attack conditions. On the whole, these analyses of existing features and machine learning algorithms together contribute to classifying the unknown test samples as genuine or spoofed. The score normalization techniques are also considered in this review to avoid any misclassifications and ultimately reduce the False Acceptance Ratios. The performance of any anti-spoofing speaker verification system may be evaluated using standard objective measures such are Equal Error Rate, False positive ratios, and graphical plots. These measures are briefly explained in this review. Overall, the critical analysis of individual methods-feature extraction, machine learning, score normalization, and all the anti-spoofing datasets are also discussed for giving a kick-start to any researcher beginning to explore in this direction. The shortcomings and risks involved in building an enhanced speaker verification system that is robust to almost all the attack types are listed in this article. The review of studies conducted so far has led to vital future directions that are enlisted in the concluding remarks of the article. Keywords: Automatic Speaker Verification; Spoofed Detection, Anti­Spoofing, Voice Conversion, Speech Synthesis, Replay Speech

A sample selection method specific to unknown test samples for calibration and validation sets based on spectra similarity

As is known to all, the construction of calibration and validation sets is of great importance for how to select representative samples into subsets so that the calibration model can be built, evaluated and predicted effectively for model development. In this study, a method was proposed for the calibration and validation sets constructed by selecting samples maximally similar to the test samples based on the spectra data. Both the Euclidean distance and Mahalanobis distance were attempted to estimate the spectra similarity. The method to select samples for calibration is more suitable and specific to unknown test samples in practical applications, thus improving the measurement accuracy. In addition, the optimization of calibration set size was carried out to avoid the influence of unnecessary samples. Two data sets of Salvia miltiorrhiza (S. miltiorrhiza) and corn by near infrared spectroscopy (NIR) were used to test the performance of the proposed method compared with two typical sample-selection algorithms, Kennard-Stone (KS) and sample set partitioning based on joint x–y distances (SPXY). The experimental results indicated that the proposed method could select a more targeted set of samples for the unknown test samples and had the superior predictive performance to the KS and SPXY methods.

A supervised blood vessel segmentation technique for digital Fundus images using Zernike Moment based features.

This paper proposes a new supervised method for blood vessel segmentation using Zernike moment-based shape descriptors. The method implements a pixel wise classification by computing a 11-D feature vector comprising of both statistical (gray-level) features and shape-based (Zernike moment) features. Also the feature set contains optimal coefficients of the Zernike Moments which were derived based on the maximum differentiability between the blood vessel and background pixels. A manually selected training points obtained from the training set of the DRIVE dataset, covering all possible manifestations were used for training the ANN-based binary classifier. The method was evaluated on unknown test samples of DRIVE and STARE databases and returned accuracies of 0.945 and 0.9486 respectively, outperforming other existing supervised learning methods. Further, the segmented outputs were able to cover thinner blood vessels better than previous methods, aiding in early detection of pathologies.

Performance evaluation of normalized difference based classifier for efficient discrimination of volatile organic compounds

Discrimination of volatile organic compounds (VOCs) is essential for mitigating their hazardous effects. Pre-processing of sensor responses and pattern recognition improves the classification results significantly. In our recent work, we reported that Normalized Difference Sensor Response Transformation (NDSRT) outperforms other popular pre-processing techniques. In this paper, we have assessed the NDSRT approach on three well-published datasets consisting of single and binary mixtures of VOCs. The raw sensor responses were transformed into virtual multi-sensor responses having concentration invariance, thereby giving rise to compacted clusters with good inter-cluster separation. In this transformation domain, even a simpler four-neuron artificial neural network (ANN) classifier was sufficient to discriminate all the test samples accurately. After application of NDSRT, Dataset—I showed a minimum cluster separation of 0.80 units while individual clusters were compacted by at least 85.01%. For Dataset—II, the average inter-cluster separation was improved by 1.4% while classification accuracy was improved from 33% to 100%. Further, for Dataset—III, the inter-cluster separation for ‘worst scenario’ was improved by at least 19% while for ‘best scenario’ it improved by 34%. Further, a simpler four neuron feed-forward back-propagation neural network (BPNN) was then used to classify 16, 30, and 50, unknown test samples taken from Dataset—I, II and III, respectively, which were not used during the training of the BPNN. The mean squared error (MSE) for these test samples was 6.7 × 10−3, 1.7 × 10−1 and 1.0 × 10−3, respectively. All the test samples were classified accurately in the proposed transformation domain.

Quantification of plasmid DNA standards for U.S. EPA fecal indicator bacteria qPCR methods by droplet digital PCR analysis

An obstacle to establishing widely useful data acceptance criteria for U.S. Environmental Protection Agency (EPA) qPCR methods has been the unavailability of standardized reference materials. Earlier versions of EPA Methods 1609 and 1611 for enterococci used cellular reference materials for quantifying enterococci in unknown test samples, however, EPA updates to these fundamentally DNA-based analysis methods have shifted toward the use of DNA standards. This report describes the application of droplet digital PCR (ddPCR) analysis for the quantification of a set of synthetic plasmid DNA standards that have been made available for updated EPA Methods 1609.1 and 1611.1 as well as for EPA Draft Method C for Escherichia coli. To obtain the most accurate concentration estimates possible, part of this effort was to develop a data analysis model for determining the fluorescence thresholds that distinguish positive from negative droplets produced by the ddPCR reactions. Versions of this model are described for applications with individual reactions, multiple reactions within a ddPCR system run, and multiple reactions within and across different system runs. The latter version was applied toward determinations of error in the concentration estimates of the standards from replicate analyses of each standard in multiple ddPCR system runs. Mean concentration estimates for the five standards from the ddPCR analyses were 4.356, 3.381, 2.371, 1.641 and 1.071 log10 copies/5 μL with associated standard deviations of 0.074, 0.082, 0.108, 0.131 and 0.188, respectively. These estimates contrasted with expected log10 concentrations of 4.6, 3.6, 2.6, 1.9 and 1.3 copies/5 μL, respectively, based on the yield of the plasmid reported by the vendor and spectrophotometric analysis of the initial stock solution of this material. These results illustrate how the analyses of original stocks may lead to potential bias(es) in the concentration estimates of final DNA standards and subsequently in the estimates of unknown test samples determined from these standards in qPCR analyses.

Quantitative characterization of bovine serum albumin thin-films using terahertz spectroscopy and machine learning methods.

The development of new spectral analysis methods in bio thin-film detection has generated intense interest in terahertz (THz) spectroscopy and its application in a wide range of fields. In this paper, it is the first time that machine learning methods are applied to the quantitative characterization of bovine serum albumin (BSA) deposited thin-films detected by terahertz time-domain spectroscopy. The spectra data of BSA thin-films prepared by solutions with concentrations ranging from 0.5 to 35 mg/ml are analyzed using the support vector regression method to learn the underlying model of the frequency against the target concentration. The learned mode successfully predicts the concentrations of the unknown test samples with a coefficient of determination R2 = 0.97932. Furthermore, aiming to identify the relevance of each frequency to the concentration, the maximal information coefficient statistical analysis is used and the three most discriminating frequencies in THz frequency are identified at 1.2, 1.1 and 0.5 THz respectively, which means a good prediction for BSA concentration can be achieved by using the top three relevant frequencies. Moreover, the top discriminating frequencies are in good agreement with the frequencies predicted by a long-wavelength elastic vibration model for BSA protein.

Direct Competitive Enzyme-Linked Immunosorbent Assay (ELISA).

The competitive enzyme-linked immunosorbent assay (ELISA) (cELISA; also called an inhibition ELISA) is designed so that purified antigen competes with antigen in the test sample for binding to an antibody that has been immobilized in microtiter plate wells. The same concept works if the immobilized molecule is antigen and the competing molecules are purified labeled antibody versus antibody in a test sample. Direct cELISAs incorporate labeled antigen or antibody, whereas indirect assay configurations use reporter-labeled secondary antibodies. The cELISA is very useful for determining the concentration of small-molecule antigens in complex sample mixtures. In the direct cELISA, antigen-specific capture antibody is adsorbed onto the microtiter plate before incubation with either known standards or unknown test samples. Enzyme-linked antigen (i.e., labeled antigen) is also added, which can bind to the capture antibody only when the antibody's binding site is not occupied by either the antigen standard or antigen in the test samples. Unbound labeled and unlabeled antigens are washed away and substrate is added. The amount of antigen in the standard or the test sample determines the amount of reporter-labeled antigen bound to antibody, yielding a signal that is inversely proportional to antigen concentration within the sample. Thus, the higher the antigen concentration in the test sample, the less labeled antigen is bound to the capture antibody, and hence the weaker is the resultant signal.

Evaluation of single and multi-feedstock biodiesel – diesel blends using GCMS and chemometric methods

Single and multi-feedstock biodiesel-diesel blends were evaluated using gas chromatography mass spectrometry along with several unsupervised and supervised chemometric methods. Peak areas of diesel alkane components and/or biodiesel fatty acid methyl ester (FAME) components were evaluated using Principal Component Analysis (PCA), k nearest neighbors (kNN), soft independent modeling of class analogy (SIMCA), and partial least squares (PLS) analysis. Using PCA (an unsupervised method), samples clustered based on feedstock type (soybean, waste grease, canola, tallow) and concentration (diesel, B2-B100). Using the supervised chemometric methods, feedstock type and concentration were validated for the training set and predicted for several unknown test samples. Concentration and feedstock were predicted using kNN, while concentration alone was predicted using SIMCA. PLS also allowed prediction of concentration but the success of the prediction heavily depended on the training model used. In addition, multi-feedstock fuel blends created from 2 and 3 feedstock components (soybean, canola, tallow) were evaluated with PCA, kNN, and SIMCA. Samples clustered based on concentration and feedstock makeup/ratio in PCA. Using kNN and SIMCA, multifeedstock blends were predicted based on concentration, while kNN could be used to predict relative ratio of multiple feedstocks. The results demonstrate the utility of chemometric analysis on a complex fuel-based data set, using methods that could be performed in a variety of laboratories and fields without the need for complex data preprocessing.

Multivariate statistical methods for detection of spur gear faults

In this study, multivariate statistical techniques are experimented for a spur gear system and a methodology is proposed. The approach is based on the analysis of multidimensional gear vibration data without any feature extradiction and data transformation. The scheme is performed using the vibration signals acquired from a lab-scale single stage gearbox in three dimensions of x, y and z directions. As a groundwork, multi-normality assumptions are established using homogeneity, autocorrelation, and univariate normality tests. The bi-dimensional frequency histograms are also plotted to show bi-normality for experimental gear data. Then, mean vectors and covariance matrices of conditions of good, worn, 1-tooth broken from wheel gear and 1-tooth broken from each pinion and wheel gear are estimated. To compare gear conditions statistically, multivariate analysis of variance is proposed and applied. Moreover, the single metric of Mahalanobis distances are calculated to classify unknown test samples, utilizing the maximum likelihood estimates. The numerical results indicate that multivariate statistical analysis techniques can be applied in early detection of spur gear faults, in which univariate tests fail.