Enrique López Droguett

Computing confidence and prediction intervals of industrial equipment degradation by bootstrapped support vector regression

Data-driven learning methods for predicting the evolution of the degradation processes affecting equipment are becoming increasingly attractive in reliability and prognostics applications. Among these, we consider here Support Vector Regression (SVR), which has provided promising results in various applications. Nevertheless, the predictions provided by SVR are point estimates whereas in order to take better informed decisions, an uncertainty assessment should be also carried out. For this, we apply bootstrap to SVR so as to obtain confidence and prediction intervals, without having to make any assumption about probability distributions and with good performance even when only a small data set is available. The bootstrapped SVR is first verified on Monte Carlo experiments and then is applied to a real case study concerning the prediction of degradation of a component from the offshore oil industry. The results obtained indicate that the bootstrapped SVR is a promising tool for providing reliable point and interval estimates, which can inform maintenance-related decisions on degrading components.

Deep learning enabled fault diagnosis using time-frequency image analysis of rolling element bearings

Traditional feature extraction and selection is a labor-intensive process requiring expert knowledge of the relevant features pertinent to the system. This knowledge is sometimes a luxury and could introduce added uncertainty and bias to the results. To address this problem a deep learning enabled featureless methodology is proposed to automatically learn the features of the data. Time-frequency representations of the raw data are used to generate image representations of the raw signal, which are then fed into a deep convolutional neural network (CNN) architecture for classification and fault diagnosis. This methodology was applied to two public data sets of rolling element bearing vibration signals. Three time-frequency analysis methods (short-time Fourier transform, wavelet transform, and Hilbert-Huang transform) were explored for their representation effectiveness. The proposed CNN architecture achieves better results with less learnable parameters than similar architectures used for fault detection, including cases with experimental noise.

Stress‐Strength Reliability Analysis with Extreme Values based on q‐Exponential Distribution

When dealing with practical problems of stress–strength reliability, one can work with fatigue life data and make use of the well-known relation between stress and cycles until failure. For some materials, this kind of data can involve extremely large values. In this context, this paper discusses the problem of estimating the reliability index R = P(Y < X) for stress–strength reliability, where stress Y and strength X are independent q-exponential random variables. This choice is based on the q-exponential distributions capability to model data with extremely large values. We develop the maximum likelihood estimator for the index R and analyze its behavior by means of simulated experiments. Moreover, confidence intervals are developed based on parametric and nonparametric bootstrap. The proposed approach is applied to two case studies involving experimental data: The first one is related to the analysis of high-cycle fatigue of ductile cast iron, whereas the second one evaluates the specimen size effects on gigacycle fatigue properties of high-strength steel. The adequacy of the q-exponential distribution for both case studies and the point and interval estimates based on maximum likelihood estimator of the index R are provided. A comparison between the q-exponential and both Weibull and exponential distributions shows that the q-exponential distribution presents better results for fitting both stress and strength experimental data as well as for the estimated R index. Copyright

A prognostics approach based on the evolution of damage precursors using dynamic Bayesian networks

During the lifetime of a component, microstructural changes emerge at its material level and evolve through time. Classical empirical degradation models (e.g. Paris Law in fatigue crack growth) are usually established based on monitoring and estimating well-known direct damage indicators such as crack size. However, by the time the usual inspection techniques efficiently identify such damage indicators, most of the life of the component would have been expended, and usually it would be too late to save the component. Therefore, it is important to detect damage at the earliest possible time. This article presents a new structural health monitoring and damage prognostics framework based on evolution of damage precursors representing the indirect damage indicators, when conventional direct damage indicator, such as a crack, is unobservable, inaccessible, or difficult to measure. Dynamic Bayesian network is employed to represent all the related variables as well as their causal or correlation relationships. Since the degradation model based on damage precursor evolution is not fully recognized, the methodology needs to be capable of online-learning the degradation process as well as estimating the damage state. Therefore, the joint particle filtering technique is implemented as an inference method inside the dynamic Bayesian network to assess both model parameters and damage states simultaneously. The proposed framework allows the integration of any related sources of information in order to reduce the inherent uncertainties. Incorporating different types of evidences in dynamic Bayesian network entails advance techniques to identify and formulate the possible interaction between potentially nonhomogenous variables. This article uses the support vector regression in order to define generally unknown nonparametric and nonlinear correlation between the input variables. The methodology is successfully applied to damage estimation and prediction of crack initiation in a metallic alloy under fatigue. The proposed framework is intended to be general and comprehensive so that it can be implemented in different applications.

AE entropy for detection of fatigue crack initiation and growth

In this paper we investigate whether the entropy of acoustic emissions (AE), can be used for determination of fatigue crack initiation and growth in AA7075-T6. The AE used for studying the damage evolution is acquired during fatigue testing of notched specimen under uniaxial cyclic load. The results show that the evolutions of the AE entropy estimated in the time domain correlates well with the evolution of the AE hit count rate. One advantage of using AE entropy is that information of both the amplitude and the compactness of AE signals is accounted for. We propose a new technique for detection and determination of fatigue crack initiation and growth, which could have a major contribution to the structural diagnosis and prognosis. We also conceptually discuss a possible connection between the thermodynamic and information entropic approach. This hypothesis may offer a novel approach to link fundamental physics of failure (PoF) with fatigue damage.

Impact Location and Quantification on an Aluminum Sandwich Panel Using Principal Component Analysis and Linear Approximation with Maximum Entropy

To avoid structural failures it is of critical importance to detect, locate and quantify impact damage as soon as it occurs. This can be achieved by impact identification methodologies, which continuously monitor the structure, detecting, locating, and quantifying impacts as they occur. This article presents an improved impact identification algorithm that uses principal component analysis (PCA) to extract features from the monitored signals and an algorithm based on linear approximation with maximum entropy to estimate the impacts. The proposed methodology is validated with two experimental applications, which include an aluminum plate and an aluminum sandwich panel. The results are compared with those of other impact identification algorithms available in literature, demonstrating that the proposed method outperforms these algorithms.

Low-frequency rTMS over the Parieto–frontal network during a sensorimotor task: The role of absolute beta power in the sensorimotor integration

Several studies have demonstrated that Repetitive Transcranial Magnetic Stimulation (rTMS) promotes alterations in the Central Nervous System circuits and networks. The focus of the present study is to examine the absolute beta power patterns in the Parieto-frontal network. We hypothesize that rTMS alters the mechanisms of the sensorimotor integration process during a visuomotor task. Twelve young healthy volunteers performed a visuomotor task involving decision making recorded (Catch a ball in a free fall) by Electroencephalography. rTMS was applied on the Superior Parietal Cortex (SPC; Brodmann area [BA] 7) with low-frequency (1 Hz - 15 min - 80% Resting Motor Threshold). For each Frontal and Parietal region, a two-way ANOVA was used to compare the absolute beta power before and after TMS for each condition of the study (Rest 1, Task and Rest 2). The results demonstrated interactions (TMS vs. Condition) for the Frontal electrodes: Fp1, Fp2 and F7 and an effect of TMS (before and after) for F4.The results for the Parietal region showed a main effect of Condition for the P3, PZ and P4 electrodes. Thus, our paradigm was useful to better understand the reorganization and neural plasticity mechanisms in the parieto-frontal network during the sensorimotor integration process.

Dynamic Bayesian Network-Based Risk Assessment for Arctic Offshore Drilling Waste Handling Practices

The increased complexity of Arctic offshore drilling waste handling facilities, coupled with stringent regulatory requirements such as zero “hazardous” discharge, calls for rigorous risk management practices. To assess and quantify risks from offshore drilling waste handling practices, a number of methods and models are developed. Most of the conventional risk assessment approaches are, however, broad, holistic, practical guides or roadmaps developed for off-the-shelf systems, for non-Arctic offshore operations. To avoid the inadequacies of traditional risk assessment approaches and to manage the major risk elements connected with the handling of drilling waste, this paper proposes a risk assessment methodology for Arctic offshore drilling waste handling practices based on the dynamic Bayesian network (DBN). The proposed risk methodology combines prior operating environment information with actual observed data from weather forecasting to predict the future potential hazards and/or risks. The methodology continuously updates the potential risks based on the current risk influencing factors (RIF) such as snowstorms, and atmospheric and sea spray icing information. The application of the proposed methodology is demonstrated by a drilling waste handling scenario case study for an oil field development project in the Barents Sea, Norway. The case study results show that the risk of undesirable events in the Arctic is 4.2 times more likely to be high (unacceptable) environmental risk than the risk of events in the North Sea. Further, the Arctic environment has the potential to cause high rates of waste handling system failure; these are between 50 and 85%, depending on the type of system and operating season. [DOI: 10.1115/1.4033713]

Risk-Based Cost-Effectiveness Analysis of Waste Handling Practices in the Arctic Drilling Operation

Research Council of Norway ENI Norge AS through the EWMA (Environmental Waste Management) project

Fully Adaptive Particle Filtering Algorithm for Damage Diagnosis and Prognosis

A fully adaptive particle filtering algorithm is proposed in this paper which is capable of updating both state process models and measurement models separately and simultaneously. The approach is a significant step toward more realistic online monitoring or tracking damage. The majority of the existing methods for Bayes filtering are based on predefined and fixed state process and measurement models. Simultaneous estimation of both state and model parameters has gained attention in recent literature. Some works have been done on updating the state process model. However, not many studies exist regarding an update of the measurement model. In most of the real-world applications, the correlation between measurements and the hidden state of damage is not defined in advance and, therefore, presuming an offline fixed measurement model is not promising. The proposed approach is based on optimizing relative entropy or Kullback–Leibler divergence through a particle filtering algorithm. The proposed algorithm is successfully applied to a case study of online fatigue damage estimation in composite materials.