Novelty Detection Methods Research Articles

Unsupervised Data-Driven Approach for Scour Damage Detection in Offshore Monopile Foundations

Scour erosion around monopile foundations significantly impacts the structural integrity and operational performance of offshore wind turbines (OWTs). To deal with this problem, this paper proposes an innovative unsupervised data-driven methodology to detect foundation scour in OWTs. The proposed method employs acceleration data collected from the tower. A numerical model of the NERL 5 MW OWT installed in medium-dense sand is developed using OpenFAST software. The study employs the novelty detection method, where only data obtained from the healthy conditions are used for training and establishing a base model. In addition, the influence of operational conditions, e.g. different wind and wave parameters is also evaluated. The process consists of three steps: (a) data pre-processing to extract time domain features and isolate environmental variations, (b) training an unsupervised anomaly detection algorithm by employing the extracted features and (c) conducting real-time anomaly detection to assess the scour damage. The study results highlight the robustness and efficiency of the proposed method in detecting scour damage around monopiles under different environmental conditions. Keywords: Monopile, scour, unsupervised, Machine learning, damage detection, SHM.

A novelty detection method for efficient data storage in smart grids

This work presents a new novelty detection method for efficiently storing electrical power system data. The proposed method makes innovative use of the Cycle-by-Cycle Difference (CCD) and the Mathematical Morphology (MM) techniques to identify novelties in the voltage of an electrical power system. Once the novelties are identified, the frames where they occur are extracted from the signal, and the other frames are discarded. Among the extracted frames, those containing power quality (PQ) disturbances are fully stored, while the frames that represent stationary states have only their main parameters stored, contributing to efficient storage. The proposed method exhibited low computational complexity in the operational phase, which was evaluated via a field-programmable gate array (FPGA) platform implementation using a customized embedded processor. The method was evaluated with real and simulated signals. The results indicate that the developed method attained high detection probabilities alongside low false alarm probabilities, resulting in a competitive Area Under the Curve (AUC) exceeding 0.976 (except for interharmonics). Comparison with other methods revealed that our approach yielded competitive results with the lowest computational cost. We concluded that the proposed method may be useful for monitoring PQ disturbances in smart grids, where the data volumes are too large.

Variational inference for semiparametric Bayesian novelty detection in large datasets

After being trained on a fully-labeled training set, where the observations are grouped into a certain number of known classes, novelty detection methods aim to classify the instances of an unlabeled test set while allowing for the presence of previously unseen classes. These models are valuable in many areas, ranging from social network and food adulteration analyses to biology, where an evolving population may be present. In this paper, we focus on a two-stage Bayesian semiparametric novelty detector, also known as Brand, recently introduced in the literature. Leveraging on a model-based mixture representation, Brand allows clustering the test observations into known training terms or a single novelty term. Furthermore, the novelty term is modeled with a Dirichlet Process mixture model to flexibly capture any departure from the known patterns. Brand was originally estimated using MCMC schemes, which are prohibitively costly when applied to high-dimensional data. To scale up Brand applicability to large datasets, we propose to resort to a variational Bayes approach, providing an efficient algorithm for posterior approximation. We demonstrate a significant gain in efficiency and excellent classification performance with thorough simulation studies. Finally, to showcase its applicability, we perform a novelty detection analysis using the openly-available Statlog dataset, a large collection of satellite imaging spectra, to search for novel soil types.

DMVSVDD: Multi-View Data Novelty Detection with Deep Autoencoding Support Vector Data Description

Novelty detection is usually defined as the identification of new or abnormal objects (outliers) from the normal ones (inliers), which has wide potential applications in the real world. Recently, an effective algorithm called Deep Support Vector Data Description (Deep SVDD) has been proposed for novelty detection, which jointly trains a deep neural network while optimizing a data-enclosing hypersphere in output space. However, some constraints such as hypersphere collapse, limit the adaptability of the model and may affect the model performance. Moreover, most of the existing studies concerning Deep SVDD focus on the novelty detection for single-view data, which may fail to provide an accurate and reliable decision because the single-view data sometimes cannot fully reflect the actual condition of the problem. In this study, we developed an end-to-end deep learning method of novelty detection for multi-view data, i.e., the Deep Multi-View SVDD (DMVSVDD). To fully preserve the correlative and complementary information of multi-view data, we jointly trained multiple deep autoencoding neural networks for multiple views while adaptively optimizing the data-enclosing hypersphere of each view in latent space. A global objective function was proposed, which takes both of the sample reconstruction error minimization and the hypersphere volume minimization into consideration simultaneously to prevent hypersphere collapse in the model. In the global objective function, the hypersphere centers and view weights of different views were designed to adaptively select the better representative features after each epoch during the training process by embedding multi-view target samples into multiple data-enclosing hyperspheres with minimum volumes. The experimental results on the MNIST-USPS and NUS-WIDE-OBJECT datasets reveal that our proposed method learns the target class effectively and is superior to some state-of-the-art methods.

Deep statistical modelling of nanopore sequencing translocation times reveals latent non-B DNA structures.

Non-canonical (or non-B) DNA are genomic regions whose three-dimensional conformation deviates from the canonical double helix. Non-B DNA play an important role in basic cellular processes and are associated with genomic instability, gene regulation, and oncogenesis. Experimental methods are low-throughput and can detect only a limited set of non-B DNA structures, while computational methods rely on non-B DNA base motifs, which are necessary but not sufficient indicators of non-B structures. Oxford Nanopore sequencing is an efficient and low-cost platform, but it is currently unknown whether nanopore reads can be used for identifying non-B structures. We build the first computational pipeline to predict non-B DNA structures from nanopore sequencing. We formalize non-B detection as a novelty detection problem and develop the GoFAE-DND, an autoencoder that uses goodness-of-fit (GoF) tests as a regularizer. A discriminative loss encourages non-B DNA to be poorly reconstructed and optimizing Gaussian GoF tests allows for the computation of P-values that indicate non-B structures. Based on whole genome nanopore sequencing of NA12878, we show that there exist significant differences between the timing of DNA translocation for non-B DNA bases compared with B-DNA. We demonstrate the efficacy of our approach through comparisons with novelty detection methods using experimental data and data synthesized from a new translocation time simulator. Experimental validations suggest that reliable detection of non-B DNA from nanopore sequencing is achievable. Source code is available at https://github.com/bayesomicslab/ONT-nonb-GoFAE-DND.

Novelty Detection and Fault Diagnosis Method for Bearing Faults Based on the Hybrid Deep Autoencoder Network

In the event of mechanical equipment failure, the fault may not belong to any known category, and existing deep learning methods often misclassify such faults into a known class, leading to erroneous fault diagnosis. In order to address the challenge of identifying new types of faults in mechanical equipment fault diagnosis, this paper proposes a novelty detection and fault diagnosis method for bearing faults based on a hybrid deep autoencoder network. Firstly, a hybrid deep autoencoder network with one input and two outputs was constructed. The original data were then fed into the network to obtain its low-dimensional representation and reconstructed data. By setting a threshold based on the reconstruction error, novel class faults can be detected, while known faults can be classified based on low-dimensional features. Experimental results demonstrate that the proposed method achieves a recognition accuracy of 98.59% (100%) for novel class identification (known fault classification) on the CWRU bearing dataset, 96.79% (98.53%) on the Paderborn dataset, and 84.34% (97.03%) on the MFPT dataset. Therefore, the hybrid deep autoencoder network not only accurately detects unknown types of faults but also effectively classifies known fault types, demonstrating excellent fault identification and classification capabilities.

Hinge Joints Performance Assessment of a PC Hollow Slab Bridge Based on Impact Vibration Testing

Hinge joint performance during the operation of prefabricated prestressed concrete (PC) hollow slab bridges is critical to ensure their lateral collaborative working condition and safe serviceability. Traditional performance identification of hinge joints mainly relies on manual inspection, which is inefficient and inaccurate. At the same time, existing indexes (such as acceleration and strain) can only qualitatively detect the damage to hinge joints. This study proposes a novelty detection method based on impact vibration testing to rapidly perform the quantitative assessment of the working condition of the hinge joints. The relationship between hinge joint performance and lateral load distribution (LLD) is first derived in detail by theoretical analysis. And then, the quantitative analysis of collaborative performance is converted to the identification of the LLD influence line, which is innovatively established by the lateral flexibility of the hollow slab bridge. The effectiveness of the proposed method is verified through a multibeam model using ABAQUS software, and the lateral collaborative working relationship between slabs is simulated using the connector elements. Furthermore, the LLD influence lines and hinge joints performance of a PC hollow slab Yanhu Bridge are evaluated based on the impact vibration testing with sensor lateral arrangement strategy. The detection results show that the proposed method can quickly and accurately identify the damage location and the stiffness loss of hinge joints.

Monitoring abnormal vibration and structural health conditions of an in-service structure from its SHM data

Vortex-induced vibration occurred recently in several important engineering structures in China. Very large mono-frequency periodic vibration was noted. This study reports on the analysis of recorded data from the structural health monitoring (SHM) system of a suspension bridge with the aim to develop some indicators that may be useful to evaluate if there is any abnormal vibration or change of health conditions of a structure in future. The power spectrum of measured acceleration responses is plotted in three-dimensions of time, frequency and power. Acceleration responses from 16 sensors on the target bridge deck in 233 days are analyzed. The evolution of different dynamic parameters in the period encompassing the occurrences of large vortex-induced vibration are reported. The changes in these parameters are identified to be closely associated with some temporary installations on the deck and temporary measures for vibration suppression. The changes in vibration power at different natural frequencies of the structure can be used to demonstrate the evolution of the unexpected event. The trend and abrupt changes of natural frequencies and modal components show the flow of vibration energy between different vibration modes. In addition to the intuitive three-dimension power spectral plot, two state vectors are also developed from the power spectrum. The identification of abnormal vibration and structural state can then be carried out automatically with the novelty detection method and a threshold. The analysis of the SHM acceleration responses of the bridge deck demonstrates the implementation procedure and the effectiveness of the proposed method.

Unsupervised machine and deep learning methods for structural damage detection: A comparative study

AbstractWhile many structural damage detection methods have been developed in recent decades, few data‐driven methods in unsupervised learning mode have been developed to solve the practical difficulties in data acquisition for civil infrastructures in different scenarios. To address such a challenge, this article proposes a number of improved unsupervised novelty detection methods and conducts extensive comparative studies on a laboratory scale steel bridge to examine their performances of damage detection. The key concept behind unsupervised novelty detection in this article is that only normal data from undamaged/baseline structural scenarios are required to train statistical models with these methods. Then, these trained models are used to identify abnormal testing data from damaged scenarios. To detect structural damage in the form of loosening bolts in the steel bridge, four machine‐learning methods (i.e., K‐nearest neighbors method, Gaussian mixture models, one‐class support vector machines, density peaks‐based fast clustering method) and one deep learning method using a deep auto‐encoder are selected. Meanwhile, some modifications and improvements are made to enable these methods to detect structural damage in unsupervised novelty detection mode. In their comparative studies, the advantages and disadvantages of these methods are analyzed based on their results of structural damage detection.

Manifold learning for novelty detection and its application in gesture recognition

As a state-of-the-art novelty detection method, Kernel Null Foley–Sammon Transform (KNFST) could identify multiple known classes and detect novelties from an unknown class via a single model. However, KNFST only captures the global information of the training set. The local geometrical structure is neglected. In this paper, a manifold is incorporated into KNFST to solve this issue. First, we use manifold graphs to depict the local structure for within-class scatter and total scatter. Second, the training samples from the same class are mapped into a single point in null space via null projected directions (NPDs). The proposed method can overcome the weakness of KNFST caused by ignoring local geometrical structure in the class. The experimental results on several toy and benchmark datasets show that manifold learning novelty detection (MLND) is superior to KNFST.

Novelty detection for metabolic dynamics established on breast cancer tissue using 2D NMR TOCSY spectra

Most metabolic profiling approaches focus only on identifying pre-known metabolites on NMR TOCSY spectrum using configured parameters. However, there is a lack of tasks dealing with automating the detection of new metabolites that might appear during the dynamic evolution of biological cells. Novelty detection is a category of machine learning that is used to identify data that emerge during the test phase and were not considered during the training phase. We propose a novelty detection system for detecting novel metabolites in the 2D NMR TOCSY spectrum of a breast cancer-tissue sample. We build one- and multi-class recognition systems using different classifiers such as, Kernel Null Foley-Sammon Transform, Kernel Density Estimation, and Support Vector Data Description. The training models were constructed based on different sizes of training data and are used in the novelty detection procedure. Multiple evaluation measures were applied to test the performance of the novelty detection methods. Depending on the training data size, all classifiers were able to achieve 0% false positive rates and total misclassification error in addition to 100% true positive rates. The median total time for the novelty detection process varies between 1.5 and 20 seconds, depending on the classifier and the amount of training data. The results of our novel metabolic profiling method demonstrate its suitability, robustness and speed in automated metabolic research.

Fault detection method based on an automated operating envelope during transient states for the large turbomachinery

In the energy generation business steam powered turbo-generators still play an important role in electrical power generation all over the world. Every facility using steam turbines considers them as the critical machinery. Such machines should be well-maintained, properly handled, and precisely diagnosed in order to achieve the best performance and safety. The most valued data about the technical health are collected during machine’s shut-downs and run-ups. These data are more than seldom and hard to assess without expert’s knowledge with strong theoretical background and experience. Main novelty of the paper is the automated method for novelty detection of machine’s vibration. Most proposed methods apply to smaller machines with rolling bearings, whereas we propose the method for large machines with sliding bearings, which have much different behavior. The application of the method is support of the plant maintenance staff to evaluate deviations of turbo-sets from a healthy state based on the concept which we called the Operating Envelope. The envelope is created based on the data from a vibration sensor during the transient state. In this paper we consider a single vibration sensor and only the first harmonic amplitude of this signal. To set the acceptance limits within which turbo-set’s dynamic response will be considered as acceptable, we used the cubic spline interpolation coupled with expert judgement. Beyond these limits the state of the turbo-set is considered as unhealthy, so it is an automated fault detection method. In such a case a machine should be a subject to further and deeper diagnostic analysis. The method was validated on the data from the 13K242 type (a 200 MW class turbine) steam turbine. We also proposed a set of parameters to evaluate the severity of malfunction.

A machine learning approach for monitoring Brazilian optical water types using Sentinel-2 MSI

Optical Water Type (OWT) is a useful parameter for assessing water quality changes related to different turbidity levels, trophic state and colored dissolved organic matter (CDOM) while also helpful for tuning chlorophyll-a algorithms. For this reason, interest in the satellite remote sensing of OWTs has recently increased in recent years. This study develops a machine learning method for monitoring Brazilian OWTs using the Sentinel-2 MSI, which can detect OWTs already assessed by field measurements and recognize new OWTs. The already assessed OWTs used for calibrating the machine learning algorithm are clear, moderate turbid, eutrophic turbid, eutrophic clear, hypereutrophic, CDOM richest, turbid, and very turbid waters. The classification method consists of two Support Vector Machines for classifying the known OWTs, while a novelty detection method based on sigmoid functions is used for assessing new OWTs. Results show the classification based on Sentinel-2 MSI bands simulated using field radiometric data is accurate (accuracy = 0.94). However, when radiometric errors are simulated, the accuracy significantly decreases to 0.75, 0.56, 0.45, and 0.37 as the mean absolute percent error increases to 10%, 20%, 30%, and 40%, respectively. Considering the errors retrieved when comparing the field and satellite measurements, the expected accuracy of Sentinel-2 MSI images is 0.78. In the satellite images, the novelty detection distinguishes new OWTs originated from the mixture among the known OWTs and a new OWT that was not part of the training database (clear blue waters). Two examples of time series in the Funil reservoir and the Curuai lake are used to show the applicability of monitoring OWTs. In the Funil reservoir, OWTs could indicate eutrophication and turbid changes caused by river inflow and sediment sinking. In the Curuai lake, OWTs could indicate areas susceptible to algae bloom and turbidity increases related to river inflow and particle resuspension. In the future, the proposed algorithm could be used for large-scale assessment of water quality degradation and supports rapid mitigation and recovery responses. For improving the classification accuracy, adjacency correction and more robust glint removal methods should be developed.

A two-stage Bayesian semiparametric model for novelty detection with robust prior information

Novelty detection methods aim at partitioning the test units into already observed and previously unseen patterns. However, two significant issues arise: there may be considerable interest in identifying specific structures within the novelty, and contamination in the known classes could completely blur the actual separation between manifest and new groups. Motivated by these problems, we propose a two-stage Bayesian semiparametric novelty detector, building upon prior information robustly extracted from a set of complete learning units. We devise a general-purpose multivariate methodology that we also extend to handle functional data objects. We provide insights on the model behavior by investigating the theoretical properties of the associated semiparametric prior. From the computational point of view, we propose a suitable varvec{xi }-sequence to construct an independent slice-efficient sampler that takes into account the difference between manifest and novelty components. We showcase our model performance through an extensive simulation study and applications on both multivariate and functional datasets, in which diverse and distinctive unknown patterns are discovered.

Comparison of Novelty Detection Methods for Detection of Various Rotary Machinery Faults.

Condition monitoring is an indispensable element related to the operation of rotating machinery. In this article, the monitoring system for the parallel gearbox was proposed. The novelty detection approach is used to develop the condition assessment support system, which requires data collection for a healthy structure. The measured signals were processed to extract quantitative indicators sensitive to the type of damage occurring in this type of structure. The indicator’s values were used for the development of four different novelty detection algorithms. Presented novelty detection models operate on three principles: feature space distance, probability distribution, and input reconstruction. One of the distance-based models is adaptive, adjusting to new data flowing in the form of a stream. The authors test the developed algorithms on experimental and simulation data with a similar distribution, using the training set consisting mainly of samples generated by the simulator. Presented in the article results demonstrate the effectiveness of the trained models on both data sets.

Early damage detection by an innovative unsupervised learning method based on kernel null space and peak‐over‐threshold

AbstractThis article proposes an innovative unsupervised learning method for early damage detection and long‐term structural health monitoring of civil structures under environmental variability. This method consists of three main parts including a novelty detector based on kernel null Foley–Sammon transform (KNFST), a practical approach to choosing an optimal Gaussian kernel parameter, and a probabilistic method for the threshold estimation. The crux of KNFST is to map all original samples to a kernel feature space and project the kernelized features into a single point in a null space. The proposed threshold estimation method exploits the extreme value theory, the generalized Pareto distribution, and the peak‐over‐threshold. The major contribution of this article is to propose an innovative novelty detection method by a one‐class kernel null space algorithm and a probabilistic threshold estimation approach. Dealing with the problem of environmental variations and estimating a reliable alarming threshold are the main advantages of the proposed method. The effectiveness and reliability of the proposed method are validated by the Wooden Bridge in a laboratory environment and the full‐scale Z24 Bridge. Results demonstrate that the proposed unsupervised learning method highly succeeds in detecting damage even under strong environmental variations.

Using internal context to detect automotive controller area network attacks

The rise in data use within cars has led to concerns about their cybersecurity. The Controller Area Network (CAN) enables communication between components core to the car’s safety and performance, and has been demonstrated to be particularly vulnerable to hacking and malicious cyber-intrusion. CAN intrusion detection systems have been envisaged. Signatures of known attacks might be used for detection, but this method holds many limitations. Although some attacks might change packet broadcast rates or add unknown packets onto the network, attacks that have little or no effect on these, yet can alter the packet data, have also been devised. We therefore test three novelty detection methods (Local Outlier Factor, Compound Classifier and One-Class Support Vector Machine) that might identify an attack based solely on anomalies in CAN packet field data-values. The methods compare values across a cluster of CAN packets broadcast from different control units, so potentially could identify an attacked control unit even when its subsequent fabricated payload data-values remain plausible. We test the methods on data from two different makes of car across a range of manipulation magnitudes, reflecting the unpredictability of attacks. Different training regimes are tested, enabling us to assess validity across journeys. We also consider the processes needed to determine the CAN fields that might be included in the intrusion detection cluster, and present algorithms for automating those processes.

Novelty detection of cable-stayed bridges based on cable force correlation exploration using spatiotemporal graph convolutional networks

The novelty detection of bridges using monitoring data is an effective technique for diagnosing structural changes and possible damages, providing a critical basis for assessing the structural states of bridges. As cable forces describe the state of cable-stayed bridges, a novelty detection method was developed in this study using spatiotemporal graph convolutional networks by analysing spatiotemporal correlations among cable forces determined from different cable dynamometers. The spatial dependency of the sensor network was represented as a directed graph with cable dynamometers as vertices, and a graph convolutional network with learnable adjacency matrices was used to capture the spatial dependency of the locally connected vertices. A one-dimensional convolutional neural network was operated along the time axis to capture the temporal dependency. Sensor faults and structural variations could be distinguished based on the local or global anomalies of the spatiotemporal model parameters. Faulty sensors were detected and isolated using weighted adjacency matrices along with diagnostic indicators of the model residuals. After eliminating the effect of the sensor fault, the underlying variations in the state of the cable-stayed bridge could be determined based on the changing data patterns of the spatiotemporal model. The application of the proposed method to a long-span cable-stayed bridge demonstrates its effectiveness in sensor fault localization and structural variation detection.

Square‐wave Anodic Stripping Voltammetric Method for Novelty Detection of Bismuth Extracted by Aqueous Two‐phase Systems

AbstractIn this work, a voltammetric sensor was used to monitor the concentration of bismuth extracted from an eutectic alloy of BiSn by aqueous two‐phase system. This strategy is a sustainable and economically viable way of recovering bismuth from secondary sources. In the aqueous two‐phase system (ATPS), biodegradable and non‐toxic constituents dispersed in water (major constituent) are used. For monitoring the extraction, bismuth was determined in the upper phase of the aqueous two‐phase system, which is rich in L35 copolymer, that causes attenuation of the electrochemical signal (anodic peak current). The electrode and operational parameters of the square‐wave anodic stripping voltammetry (SWASV) were evaluated according to the deposition and stripping processes of the bismuth on the surface of the carbon paste electrode (CPE). It was observed a similarity between the electrochemical response of the bismuth extracted by ATPS and with the standard solution of bismuth. The proposed method shows a linear range of 1.29–8.94 μmol L−1, limit of detection (1.07 μmol L−1) and limit of quantification (3.57 μmol L−1) and good precision (RSD%=2.27 %). This method was validated by comparing the results with Flame Atomic Absorption Spectroscopy (FAAS), using statistical tests to verify precision and accuracy. In conclusion, using a voltammetric sensor to monitor the concentration of bismuth extracted by ATPS proved to be an efficient method, in agreement with the concentrations of the referenced method.

Zero-Shot Learning for EEG Classification in Motor Imagery-Based BCI System.

A brain-computer interface (BCI) based on motor imagery (MI) translates human intentions into computer commands by recognizing the electroencephalogram (EEG) patterns of different imagination tasks. However, due to the scarcity of MI commands and the long calibration time, using the MI-based BCI system in practice is still challenging. Zero-shot learning (ZSL), which can recognize objects whose instances may not have been seen during training, has the potential to substantially reduce the calibration time. Thus, in this context, we first try to use a new type of motor imagery task, which is a combination of traditional tasks and propose a novel zero-shot learning model that can recognize both known and unknown categories of EEG signals. This is achieved by first learning a non-linear projection from EEG features to the target space and then applying a novelty detection method to differentiate unknown classes from known classes. Applications to a dataset collected from nine subjects confirm the possibility of identifying a new type of motor imagery only using already obtained motor imagery data. Results indicate that the classification accuracy of our zero-shot based method accounts for 91.81% of the traditional method which uses all categories of data.