Punching process monitoring using wavelet transform based feature extraction and semi-supervised clustering

The failure detection and location method (FDLM) is discussed. The key idea is to utilize the causal description of the system's power flows. This description is used to detect and locate system failures without modeling the possible system failure modes. The FDLM specifies those system and model outputs the comparison of which maximize the possibilities of detecting existing system failures. An example of an FDLM application in a simulated fossil fuel power plant demonstrates that the method is applicable to systems which are modeled by bilateral as well as unilateral and multilateral energy bonds. An integration of the FDLM with other available failure detection methods seems to result in an effective and promising failure detection scheme.

The driver's cognitive and physiological states affect his/her ability to control the vehicle. Thus, these driver states are essential to the safety of automobiles. The design of advanced driver assistance systems (ADAS) or autonomous vehicles will depend on their ability to interact effectively with the driver. A deeper understanding of the driver state is, therefore, paramount. Electroencephalography (EEG) is proven to be one of the most effective methods for driver state monitoring and human error detection. This paper discusses EEG-based driver state detection systems and their corresponding analysis algorithms over the last three decades. First, the commonly used EEG system setup for driver state studies is introduced. Then, the EEG signal preprocessing, feature extraction, and classification algorithms for driver state detection are reviewed. Finally, EEG-based driver state monitoring research is reviewed in-depth, and its future development is discussed. It is concluded that the current EEG-based driver state monitoring algorithms are promising for safety applications. However, many improvements are still required in EEG artifact reduction, real-time processing, and between-subject classification accuracy.

All failure detection methods are based, either explicitly or implicitly, on the use of redundancy, i.e. on (possibly dynamic) relations among the measured variables.The robustness of the failure detection process consequently depends to a great degree on the reliability of the redundancy relations, which in turn is affected by the inevitable presence of model uncertainties.In this paper we address the problem of determining redundancy relations that are optimally robust, in a sense that includes several major issues .ofimportance in practical failure detection, and that provides a significant amount of intuition concerning the geometry of robust failure detection.We also give a procedure, involving the construction of a single matrix and its singular value decomposition, for the determination of a complete sequence of redundancy relations, ordered in terms of their level of robustness.This procedure also provides the basis for comparing levels of robustness in redundancy provided by different sets of sensors.

This paper determines the theoretical limits of failure detectability in systems with modeling errors. Such errors are the main obstacle to accurate failure detection in aircraft, jet engines, power generation systems, and most other systems that include mechanical components and/or thermofluid processes. The paper then introduces the reachable measurement intervals (RMI) failure detection method that performs at the limit of detectability. RMI is capable of detecting the smallest theoretically detectable failures in systems with modeling errors, and is implementable in real time on microprocessors. The method is illustrated on a jet engine example. Nomenclature a — entry of the matrix A A = system matrix b = entry of the matrix B B = input matrix Bj = input matrix equal to the identity matrix C = output matrix C{ = output matrix for additional measurements D = direct transmission matrix H = Pontryagin's function K — matrix of gains for additional measurements L = vector of Lagrange multipliers m = entry of the matrix M M = matrix of Lagrange multipliers associated with uncertain entries of A n = entry of the matrix N N = matrix of Lagrange multipliers associated with uncertain entries of B p = discrete-time index corresponding to current time P = performance index for optimization q - vector of system parameters / = time t0 — current time ts = length of the processing window T - sampling time u = input vector x = state vector y = output vector z - vector of additional measurements Subscripts d = discrete e = uncertain and uncorrelated n = nominal v = uncertain

Recent advances in microelectronic circuit have enabled the application of process diagnostics to a variety of systems to improve performances and the reliability. Failure detection and isolation strategies monitor a system for degradations and if detected, classify the failure source. One of the most important methods for failure detection and diagnosis is the analysis of the variations of the estimated parameter of DC Motor. In this paper is presented a structure of a method for failure detection and diagnosis based on continuous time parameter estimation of a mathematical model of the DC Motor. Continuous time parameters can be related easily to the physical characterizations of DC Motor. The parameter is directly estimated by well-known prediction error method of identification for dynamic systems. The failure detection it is based on changing the process coefficient and statistical decision methods. Unlike usual papers, the failure can be detected whether others process parameters are not constant.

Among the fastest‐growing applications of high‐precision GPS positioning are those which are kinematic in nature. Carrier phase‐based GPS positioning of a moving antenna—for example, attached to a ship, an aircraft, or a land vehicle—is now commonplace. Recent software innovations make use of advanced ambiguity resolution “on the fly” and real‐time kinematic data processing algorithms to emulate the ease of operation of conventional differential GPS (DGPS) based on transmitted pseudo‐range corrections. However, as much higher accuracy must now be assured compared to DGPS, greater attention must be focused on the quality control aspects of GPS positioning. This study describes two methods for detecting failures or changes of small magnitude in real time in GPS measurements. Examination of the overlap or disjointedness of robust and conventional confidence intervals and studentized normal variates have been used as failure detection tools. These methods are based on testing the performance of the differences between the conventional (nonrobust) Kalman state estimates and the robust Kalman filler estimates. Detection of cycle slips in carrier phase data, outliers in phase rate or in code ranges, or any other type of disorder in the measurements of the GPS system can be addressed with these failure detection methods. Application and evaluation of the algorithms has been carried out using raw carrier‐phase and phase‐rate GPS measurements. It has been demonstrated that these failure detection tools provide powerful and efficient diagnostics for detecting small changes in the measurements of the GPS system.

Failure detection methods are of significant interest for photovoltaic (PV) site operators to help reduce gaps between expected and observed energy generation. Current approaches for field-based fault detection, however, rely on multiple data inputs and can suffer from interpretability issues. In contrast, this work offers an unsupervised statistical approach that leverages hidden Markov models (HMM) to identify failures occurring at PV sites. Using performance index data from 104 sites across the United States, individual PV-HMM models are trained and evaluated for failure detection and transition probabilities. This analysis indicates that the trained PV-HMM models have the highest probability of remaining in their current state (87.1% to 93.5%), whereas the transition probability from normal to failure (6.5%) is lower than the transition from failure to normal (12.9%) states. A comparison of these patterns using both threshold levels and operations and maintenance (O&M) tickets indicate high precision rates of PV-HMMs (median = 82.4%) across all of the sites. Although additional work is needed to assess sensitivities, the PV-HMM methodology demonstrates significant potential for real-time failure detection as well as extensions into predictive maintenance capabilities for PV.

In order to take advantage of the merits of WPT and HHT in feature extraction from vibration signals of power transformer, a time-scale-frequency analysis method is developed based on the combination of these two techniques. This method consists of two steps. First, the desirable wavelet packet nodes corresponding to characteristic frequency bands of power transformer are selected through a Correlation Degree Threshold Screening (CDTS) technique for reconstructing a time-domain signal that contains useful information of power transformer. Second, the HHT is then conducted on the reconstructed signal to track the instantaneous frequencies corresponding to natural characteristics of power transformer. Experimental results are provided by analyzing a real power transformer vibration signal. Compared with the features extracted by directly using HHT, the features obtained by the proposed method reveal clearer condition pattern of the transformer, which shows the potential of this method in condition monitoring of power transformer.

Mammogram—breast X-ray—is considered the most effective, low cost, and reliable method in early detection of breast cancer. Although general rules for the differentiation between benign and malignant breast lesions exist, only 15–30 % of masses referred for surgical biopsy are actually malignant. In this work, an approach is proposed to develop a computer-aided classification system for cancer detection from digital mammograms. The proposed system consists of three major steps. The first step is region of interest (ROI) extraction of 256 × 256 pixels size. The second step is the feature extraction; we used a set of 19 GLCM and GLRLM features, and the 19 (nineteen) features extracted from gray-level run-length matrix and gray-level co-occurrence matrix could distinguish malignant masses from benign masses with an accuracy of 96.7 %. Further analysis was carried out by involving only 12 of the 19 features extracted, which consists of 5 features extracted from GLCM matrix and 7 features extracted from GLRL matrix. The 12 selected features are as follows: Energy, Inertia, Entropy, Maxprob, Inverse, SRE, LRE, GLN, RLN, LGRE, HGRE, and SRLGE; ARM with 12 features as prediction can distinguish malignant mass image and benign mass with a level of accuracy of 93.6 %. Further analysis showed that area under the receiver operating curve was 0.995, which means that the accuracy level of classification is good or very good. Based on that data, it was concluded that texture analysis based on GLCM and GLRLM could distinguish malignant image and benign image with considerably good result. The third step is the classification process; we used the technique of decision tree using image content to classify between normal and cancerous masses. The proposed system was shown to have the large potential for cancer detection from digital mammograms.

Recent years have witnessed impressive robotic manipulation systems driven by advances in imitation learning and generative modeling, such as diffusion-and flow-based approaches.As robot policy performance increases, so does the complexity and time horizon of achievable tasks, inducing unexpected and diverse failure modes that are difficult to predict a priori.To enable trustworthy policy deployment in safety-critical human environments, reliable runtime failure detection becomes important during policy inference.However, most existing failure detection approaches rely on prior knowledge of failure modes and require failure data during training, which imposes a significant challenge in practicality and scalability.In response to these limitations, we present FAIL-Detect, a modular two-stage approach for failure detection in imitation learning-based robotic manipulation.To accurately identify failures from successful training data alone, we frame the problem as sequential out-of-distribution (OOD) detection.We first distill policy inputs and outputs into scalar signals that correlate with policy failures and capture epistemic uncertainty.FAIL-Detect then employs conformal prediction (CP) as a versatile framework for uncertainty quantification with statistical guarantees.Empirically, we thoroughly investigate both learned and post-hoc scalar signal candidates on diverse robotic manipulation tasks.Our experiments show learned signals to be mostly consistently effective, particularly when using our novel flow-based density estimator.Furthermore, our method detects failures more accurately and faster than state-of-the-art (SOTA) failure detection baselines.These results highlight the potential of FAIL-Detect to enhance the safety and reliability of imitation learning-based robotic systems as they progress toward realworld deployment.Videos of FAIL-Detect can be found on our website: https://cxu-tri.github.io/FAIL-Detect-Website/.

The Fourier transform has been widely used in radar signal and image processing. When the radar signals exhibit time- or frequency-varying behavior, an analysis that can represent the intensity or energy distribution of signals in the joint time-frequency (JTF) domain is most desirable. In this article, we showed that JTF analysis is a useful tool for improving radar signal and image processing for time- and frequency-varying cases. We applied JTF analysis to radar backscattering and feature extraction; we also examined its application to radar imaging of moving targets. Most methods of JTF analysis are non-parametric. However, parametric or model-based methods of time-frequency analysis, such as adaptive Gaussian and chirplets, are more suitable for radar signals and images.

With the increasing complexity of software-intensive systems, software health management is proposed to assure their runtime dependability, in which online failure detection and prediction is one of the most significant components. Failure indicators are characteristics of internal states and behavior of a system which indicate potential failures. However, previous studies mostly focused on extracting failure indicators from network and hardware outside of a software system or operating system level, neglected the runtime dynamics on application level. Besides, most of these studies aimed at detecting and predicting performance-related failures. As a major category of software failures, content failures are often omitted. This paper proposes an experiment-based approach to extract interaction-related failure indicators on application level for content failures, composed of abnormal execution time of modules and abnormal interaction times between modules. Firstly, an experiment-based failure data generation method is proposed due to a lack of real-world failure data which can reflect the runtime states and behavior of a software system. Then a machine learning method is selected and applied on the failure dataset to construct classifiers for normal data and failure data, from which failure indicators are extracted. Finally, three open-source software were selected to show the validity of our extracting method and the effectiveness of the extracted failure indicators. Interaction-related failure indicators extracted by the proposed approach can be used for runtime failure detection and prediction of content failures, thus improving runtime dependability of complex software-intensive systems.

A general-purpose useful parameter in data analysis is the intrinsic dimension of a data set, corresponding to the minimum number of variables necessary to describe the data without significant loss of information. Feature extraction, including linear or nonlinear mapping technique, is efficient to estimate the intrinsic dimension of the data set, which is a key issue to machine fault diagnosis. This paper presents a novel application of feature extraction using the nonlinear mapping technique called curvilinear component analysis (CCA) for gear failure detection. In the approach high-dimensional data are nonlinearly projected toward an output space with dimension equal to the intrinsic dimension. Hence, enough information is remained to describe correctly the original data structure, and feature extraction based on CCA reduces dimensionality of the raw feature space for machine failure detection. Gearbox vibration signals measured under different operating conditions are analyzed using the technique. The results indicate that the intrinsic dimension of the data set is estimated and a 2-D subspace is extracted by the CCA technique, then the high-dimensional original feature data are projected into the 2-D space and form several clustering regions, each indicative of a specific gear condition, respectively. Thus, the gear operating conditions including normal, one cracked tooth, and one broken tooth are classified and detected clearly. It confirms that feature extraction based on the nonlinear mapping is very useful and effective for pattern recognition in mechanical fault diagnosis, and provides a good potential for applications in practice.

The effective condition monitoring of planetary gearboxes is important for making informed maintenance decisions and minimizing machine downtime. Traditional physics-based and data-driven approaches for condition monitoring of planetary gearboxes have inherent shortcomings that limit their practical application for remaining useful life prediction. Hybrid approaches show unique potential for performing diagnostics and prognostics of planetary gearboxes in situations where failure data is not available. Hybrid methods incorporate both physics-based and data-driven models that exploit their respective advantages. This work starts with a brief overview of the various challenges associated with vibration-based condition monitoring of planetary gearboxes and the characteristics of planetary gearboxes that complicate the condition monitoring process. Thereafter, the potential of hybrid methods to address these challenges is discussed. The respective advantages and disadvantages of physics-based and data-driven approaches are summarized and a short review of the use of hybrid methods in condition monitoring is presented. Finally, various considerations for developing hybrid methods for planetary gearbox diagnostics and prognostics are discussed. This includes considerations for selecting an appropriate hybrid framework, pre-processing of the measured vibration signal, physics-based modelling of a planetary gearbox, model calibration, health state estimation, and health state prediction.

Guaranteed Active Failure Detection and Isolation for Linear Dynamical Systems