Real-time Failure Research Articles

In-Situ Visualization of Tensile Failure in Additively Manufactured 316 L Stainless Steel

Additive manufacturing has become an enabling technology for the production of complicated engineered structures once thought impossible to produce. As these technologies develop, the mechanical behavior of these materials/structures must be characterized in a variety of harsh environments. To assess the loading-rate sensitivity of additively manufactured 316 L stainless steel, dynamic and quasi-static tension experiments were performed. High-speed X-ray phase contrast imaging was performed during dynamic experiments at Argonne National Laboratory’s Advanced Photon Source. These images reveal the evolution of porosity and intrinsic defects within the material, and their influence on the mechanisms of dynamic failure in real time. Stress-strain histories were recorded for experiments, on which the performance of the material is addressed. High degrees of localized yielding were observed as a precursor to ductile crack growth and propagation. No transition in the mechanism of fracture was observed. However, from the stress strain-histories the influence of defects was shown to decrease with increasing strain-rates. Comparisons between failure mechanisms of the additively manufactured specimens were made to those of wrought material subjected to identical loading.

Online Monitoring Machining Errors of Thin-Walled Workpiece: A Knowledge Embedded Sparse Bayesian Regression Approach

Deflection of the tool and workpiece caused by cutting forces usually leads to machining errors of thin-walled workpieces. Monitoring this kind of force-induced errors plays an extremely important role in controlling and compensating the deflection-related machining failures in real time. However, accompanied by time consuming and complicated computations, now available analytical prediction methods cannot satisfy the requirements of online machining errors prediction. Hence, data-driven regression methods are introduced to online predict the machining errors. The challenges lie in that: first, the spatial continuous distribution of machining errors needs to be constructed via limited measured points; second, the regression model must have high generalization performance to adapt varied cutting parameters; and third, the model complexity should be restrained to improve the real-time performance. To tackling these challenges, a knowledge embedded regression is presented to model the relationship between machining error and cutting parameters, cutting location, as well as online measured cutting forces. The physical mechanism about machining errors is integrated into the model for improving the generalization accuracy. A Gaussian prior distribution over the model weights is introduced to reduce the model redundancy, for the sake of learning the weight variables with limited samples and increasing the prediction efficiency. Results have indicated that the predicted machining errors by the proposed model accords with the measurement better than those predicted by a purely data-dependent regression.

Recurrent neural network based real-time failure detection of storage devices

Studies have revealed that the failure rates of storage devices can often be as high as fourteen percent. To make matters worse, there are frequently no warning signs for precaution before catastrophic failure of storage devices occurs. A real-time predictive maintenance system that provides an automatic means for predicting when maintenance should be performed to ultimately eliminate unexpected breakdowns needs to be developed. Unlike traditional regression predictive modeling, the failure detection of storage devices is a problem of time series prediction, which adds the complexity of a sequence dependence among the input variables. The proposed LSTM (Long Short-Term Memory) network is a branch of RNN (Recurrent Neural Network) used in deep learning, which presents a very large architecture that can be successfully trained. LSTM is good at extracting patterns in input feature space, where the input data spans over long sequences. With the gated architecture of LSTM, it is capable of learning the context required to make predictions in time series forecasting. It is ideal for generating responses that depend on a time-evolving state; for example detecting the condition of storage devices over time. This paper describes our development of an LSTM (Long short-term memory), a special kind of RNN (Recurrent Neural Network)—based real-time predictive maintenance system (RPMS) built on top of Apache Spark for detecting storage device failure. By streaming real-time data into a RPMS directly from the device itself, the issues can be revealed and addressed early before they cause costly downtime.

융합 컴퓨팅] 컴퓨팅 자원의 가용성을 보장하기 위한 기계 학습 기반의 실시간 장애 예측 프레임워크 연구

컴퓨터 시스템을 통해 수많은 서비스들이 제공됨에 따라 시스템의 신뢰성 및 가용성이 중요해졌다. 이에 시스템 장애를 예측하여 이를 사전에 방지하는 것이 주요 과제가 되었다. 기존 연구에서 복잡한 시스템의 가용성을 보장하기 위해, 가장 많은 피해 비용을 초래하는 시스템 구성 자원의 장애를 예측하고자 하였으며, 시스템 지식을 기반으로 최적의 데이터 가공 및 예측 알고리즘을 적용하여 높은 예측률의 모델을 얻고자 하였다. 하지만, 최적의 모델을 얻기 위해서는 반복적인 데이터 분석 및 가공, 예측 모델 최적화 및 비교가 필요하였고 해당 과정의 일부만이 경험적 지식에 의존하여 수동적으로 수행되었다. 본 논문에서는 최적화된 예측 모델을 얻기 위한 과정을 전략적으로 자동화하여 이러한 비용을 최소화하는 프레임워크를 구현하였다. 이를 위해 기존 연구에서 장애와 높은 연관성이 검증된 데이터를 수집하는 모니터링 시스템과 자동화된 기계 학습을 적용하여 특징 공학, 알고리즘 선택, 예측모델 최적화 과정을 자동화하고, 생성된 모델을 기반으로 실시간 장애 예측을 가능하게 하였다. 또한 여러 논문에서 파편화되어 있는 장애 개념, 예측 방법 및 적용을 체계화하고 프레임워크에 반영하였다.

A Real-Time Flight Safety Alarm System for Multiple UAVs

Airborne heterogeneous UAVs can encounter flight control or sensor failures while performing flight tasks. To protect against this, safety alarm systems (SAS) are typically implemented in ground control systems (GCS) to prevent crashes or other failures. Tracing the failure for a single UAV is relatively easy, but it is difficult to manually monitor multiple UAVs flying in a distributed open area. The current paper aims to fulfil this SAS deficiency, a deficiency that is still frequently ignored during GCS development. Our SAS alerts users of no-fly zones as well as navigation, battery, GPS and communication failures. The system can monitor simultaneous multiple UAV-flight failures in real time so that GCS can send appropriate commands to each UAV encountering a problem.

Parametric selection of software reliability growth models using multi-criteria decision-making approach

The present research aims at the development of a computation procedure based on the mathematical procedures to solve the problem of Software Reliability Growth Model's (SRGMs) selection problem. The present selection problem is set as a Multi-Criteria Decision Making (MCDM) problem in this study and an MCDM approach, namely VIseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR) is implemented to solve the same. To demonstrate the working of the VIKOR, two real-time failure data sets have been used on which 16 SRGMs are evaluated against a set of 12 selection indexes. The present study produces a comprehensive ranking of the various SRGMs based on their relative proximity as the result. The model having the smallest proximity value is graded at position-1 and the model with largest proximity value at the last position.

Do Not Make the Same Mistakes Again and Again: Learning Local Recovery Policies for Navigation From Human Demonstrations

In this paper, we present a human-in-the-loop learning framework for mobile robots to generate effective local policies in order to recover from navigation failures in long-term autonomy. We present an analysis of failure and recovery cases derived from long-term autonomous operation of a mobile robot, and propose a two-layer learning framework that allows to detect and recover from such navigation failures. Employing a learning by demonstration (LbD) approach, our framework can incrementally learn to autonomously recover from situations it initially needs humans to help with. The learning framework allows for both real-time failure detection and regression using Gaussian processes (GPs). Our empirical results on two different failure scenarios indicate that given 40 failure state observations, the true positive rate of the failure detection model exceeds 90%, ending with successful recovery actions in more than 90% of all detected cases.

Structure Learning in Power Distribution Networks

Traditional power distribution networks suffer from a lack of real-time observability. This complicates development and implementation of new smart-grid technologies, such as those related to demand response, outage detection and management, and improved load monitoring. In this paper, inspired by proliferation of metering technology, we discuss topology estimation problems in structurally loopy but operationally radial distribution grids from measurements, for example, voltage data, which are either already available or can be made available with a relatively minor investment. The primary objective of this paper is to learn the operational layout of the grid. Further, the structure learning algorithm is extended to cases with missing data, where available observations are limited to a fraction of the grid nodes. The algorithms are computationally efficient—polynomial in time—which is proven theoretically and illustrated in numerical experiments on a number of test cases. The techniques developed can be applied to detect line failures in real time as well as to understand the scope of possible adversarial attacks on the grid.

Real-time risk assessment of casing-failure incidents in a whole fracturing process

Abstract With the increasing development of hydraulic fracturing technologies, shale gas exploitation is becoming a highly industrial process. Casing-failure incidents which could cause serious leakage of the high-pressure and hazardous chemicals have triggered an intense public discussion in the fracturing industry. Quantitative risk analysis (QRA) is a common technique used to study the carrying capacity of a casing. However, this technique tends to only show static risk state after the casing has been run down to the formation, but is not sufficient to monitor its real-time failure risk during the whole fracturing process. Therefore, a matrix-based risk assessment method is proposed to improve the conventional QRA by using stress-strength interference theory and value function modelling to calculate the static and dynamic failure probabilities of casings, respectively, over a period of multi-stage sand fracturing. Further studies are developed to integrate these two probabilities with the application of a design matrix, particularly for the quantitative analysis and assessment of casing failures. The visual risk graphs are also provided to show the failure risk states and levels for the casings in real time. The assessment procedures can clearly delineate the operation characteristics of shale gas fracturing − high pressure, large displacement, and sand erosion. To illustrate the validity of the methodology, a production casing of a gas well at one fracturing stage is chosen as a test case. Results show that the real-time risk is more accurate and practical, as well as improving the assessment effectiveness of casing-failure incidents during a whole fracturing period.

Resilient Protection System Through Centralized Substation Protection

Power systems are experiencing drastic changes with the introduction of renewables and customer owned resources. These changes introduce new challenges in protection, control, and operation of power systems. In protection, it is necessary to develop new methods that are immune to the new characteristics and they avoid relay misoperations. In particular, any protection system is vulnerable to hidden failures. This paper introduces a new dynamic state estimation-based centralized protection scheme (DSEBCPS) at a substation level. The system supplements dynamic state estimation-based protection for individual zones known as “setting-less relays” to secure their operation against hidden failures. The DSEBCPS communicates with the setting-less relays via the station bus and obtains essential information from each protection zone, such as phasor quantities and breakers/disconnects status. This information is processed by the DSEBCPS to estimate substation topology and states. Specifically, the DSEBCPS performs dynamic state estimation in the quasi dynamic domain once per cycle to detect any abnormality within the substation. Upon detecting abnormalities, the DSEBCPS performs hypothesis testing to identify the source of the abnormality, that is, a fault or a hidden failure. Upon detecting a hidden failure, the DSEBCPS identifies the compromised data, replaces this data with estimated values by using the substation estimated states and model. The corrected data are used by the setting-less relays, enabling self-healing characteristics. The proposed capability to detect hidden failures and correct the effects of these failures in real time makes the setting-less relays secure against misoperations and bridges a critical gap in protection systems. This capability is demonstrated through numerical experiments that include several scenarios of hidden failures. The integration of the proposed scheme with the individual protection zone relays form a resilient protection and control system, which is self-immunized against hidden failures.

Concrete–fiber-reinforced polymer interfacial bond monitoring with self-triggering sensors

External bonding with fiber-reinforced polymers is currently one of the most popular technologies for rehabilitation of concrete structures. However, the effectiveness of the technology largely depends on the strength of the bond between the fiber-reinforced polymer laminate and the concrete substrate. This article provides a system to monitor the loss of bond between the fiber-reinforced polymer laminate and the concrete. Fiber optic sensors are broadly accepted as a structural health monitoring device for fiber-reinforced polymer materials by integrating the sensors into the host material. A recent development in fiber optic sensor technology is the mechanoluminescence-based optoelectronic sensors. Concrete beams strengthened with multifunctional fiber-reinforced polymer laminates were tested in shear using these sensors to evaluate the bond strength of the composite system. The sensors showed response to shear stress transfer in the adhesive layer which was observed to be as low as 2 MPa. The inclusion of sensors does not affect the bond strength (3.35 MPa), for both beams with sensors and without sensors. Real-time failure detection of fiber-reinforced polymer–strengthened beams was successfully achieved in this study. In future, the scheme aims at providing a tool to reduce the response time and decision making involved in the maintenance of deficient structures.

A Reliability Assessment of the Hydrostatic Test of Pipeline with 0.8 Design Factor in the West–East China Natural Gas Pipeline III

The use of 0.8 design factor in Chinese pipeline industry is a breakthrough with the success of the test pipe section in the west–east China gas pipeline III. For such a design factor, the traditional P-V (Pressure-Volume) curve based pressure test control cannot describe the details of the process, and the 0/1 type failure is not an efficient index to show the safety level of the pipeline. In this paper, a reliability based assessment method is proposed to monitor the real-time failure probability of the pipeline during the hydrostatic test process. The reliability index can be used as the degree of risk. Following the actual hydrostatic testing of a test pipe section with 0.8 design factor in the west–east China gas pipeline III, reliability analysis was performed using Monte Carlo technique. The basic values of input parameters of the limit state equations are based on the data collected from either the tested section or the recommended value in the codes. The analysis of limit states, i.e., the yielding deformation and the excessive plastic deformation of pipeline, proceeded based on these distributions. Finally, it is found that the gradually increased water pressure makes the failure probability increase accordingly. A reliability assessment method was proposed and illustrated with the practical pressure test process.

Contribution of AE analysis in order to evaluate time to failure of ceramic matrix composites

The paper presents an empirical approach to obtain the Remaining Useful Lifetime (RUL) or Time To Failure (TTF) using mechanical parameters or Acoustic Emission (AE) of Ceramic Matrix Composite (CMC) submitted to static fatigue tests at the laboratory scale. The approach is based on the investigation of the AE dynamic and identification of characteristic times with several indicators such as the sentry function, coefficient RAE, coefficient of attenuation B. A linear correlation is established between these characteristics times and the time to failure allowing the possibility of evaluating the Time To Failure in real time. The results are compared to those obtained with mechanical parameters such as the strain and the damage parameter.

Feeling machines for online detection and compensation of tool deflection in milling

In milling, the tool deflection negatively affects the accuracy of the workpiece geometry. Feeling machines with force sensing capabilities represent a promising enabler to detect and to compensate the deflection-related machining failures in real time. This paper gives an overview about current applications of feeling machine tools and developments of monitoring and compensation systems for tool deflection. At first, a method to integrate strain sensors within rigid machine tool structures is introduced. Secondly, the prototypical application within commercial milling centres is presented. Finally, two compensation strategies for tool deflection via feed- and position-based control are discussed and experimentally evaluated.

Failure warning of hydrous sandstone based on electroencephalogram technique

Sandstone is a type of rock mass that widely exists in nature. Moisture is an important factor that leads to sandstone structural failure. The major failure assessment methods of hydrous sandstone at present cannot satisfy real-time and portability requirements, especially lacks of warning function. In this study, acoustic emission (AE) and computed tomography (CT) techniques are combined for real-time failure assessment of hydrous sandstone. Eight visual colors for warning are screened according to different failure states, and an electroencephalogram (EEG) experiment is conducted to demonstrate their diverse excitations of the human brain’s concentration.

A real-time tool failure monitoring system based on cutting force analysis

This paper developed a novel, simple-to-use method for detecting tool failure in cutting. An EEMD filtering and restructuring method is launched according to entropy and correlation of cutting force signal in turning high-strength and high-hardness materials using carbide cutting tools. Then, friction coefficient and mean power theories are adopted to characterize tool failure status using restructured cutting force signal. Finally, an automatic detector is proposed using CUSUM control chart to monitor tool chipping and tool wear failure in real time based on previous analysis and can be verified by semi-finishing experiment. Without any need of other sensor equipment and long computing time, the method can be reliable and simple to operate in detecting tool failure in real time.

Detection of Interfacial Debonding in a Rubber-Steel-Layered Structure Using Active Sensing Enabled by Embedded Piezoceramic Transducers.

Rubber–steel-layered structures are used in many engineering applications. Laminated rubber–steel bearing, as a type of seismic isolation device, is one of the most important applications of the rubber–steel-layered structures. Interfacial debonding in rubber–steel-layered structures is a typical failure mode, which can severely reduce their load-bearing capacity. In this paper, the authors developed a simple but effective active sensing approach using embedded piezoceramic transducers to provide an in-situ detection of the interfacial debonding between the rubber layers and steel plates. A sandwiched rubber–steel-layered specimen, consisting of one rubber layer and two steel plates, was fabricated as the test specimen. A novel installation technique, which allows the piezoceramic transducers to be fully embedded into the steel plates without changing the geometry and the surface conditions of the plates, was also developed in this research. The active sensing approach, in which designed stress waves can propagate between a pair of the embedded piezoceramic transducers (one as an actuator and the other one as a sensor), was employed to detect the steel–rubber debonding. When the rubber–steel debonding occurs, the debonded interfaces will attenuate the propagating stress wave, so that the amplitude of the received signal will decrease. The rubber–steel debonding was generated by pulling the two steel plates in opposite directions in a material-testing machine. The changes of the received signal before and after the debonding were characterized in a time domain and further quantified by using a wavelet packet-based energy index. Experiments on the healthy rubber–steel-layered specimen reveal that the piezoceramic-induced stress wave can propagate through the rubber layer. The destructive test on the specimen demonstrates that the piezoceramic-based active sensing approach can effectively detect the rubber–steel debonding failure in real time. The active sensing approach is often used in structures with “hard” materials, such as steel, concrete, and carbon fiber composites. This research lays a foundation for extending the active sensing approach to damage detection of structures involving “soft” materials, such as rubber.

Implementation of a Communication Bundle for High-Risk Patients.

Interventions that facilitate early identification and management of hospitalized pediatric patients who are at risk for deterioration are associated with decreased mortality. In our large pediatric hospital with a history of success in decreasing unrecognized deterioration, patients at higher risk of deterioration are termed "watchers." Because communication errors often contribute to unrecognized deterioration, clear and timely communication of watcher status to all team members and contingency planning was desired. Increase the percentage of eligible watchers with a complete communication, teamwork, and planning bundle within 2 hours of identification from 28% to 80%. Watchers admitted to Hospital Medicine on 2 targeted units were eligible. Stakeholders were educated to facilitate ownership. Daily data analysis enabled real-time failure identification. Automated physician notification provided reminders for timely communication. The percentage of watchers with a complete situation awareness bundle within 2 hours increased from 28% to 81% and was sustained for more than 2 years. There was no change in rates of rapid response team calls or ICU transfers on our intervention units, but these both increased throughout the hospital. Education facilitated modest improvement, with marked improvements and sustainment through use of technology. A novel bundle that included contingency planning and communication expectations was created to improve situation awareness for watchers. Multidisciplinary engagement and use of automated technology facilitated by an electronic health record helped implement and sustain bundle adherence.

Evaluation to determine the efficiency for the diagnosis search formation method of failures in automated systems

This paper describes the results of work in the field of failure self-diagnostics for automated systems in real time to increase the efficiency of their operation. We describe the method developed of a diagnosis search formation space by applying to the Expert System Knowledge Base to diagnose failures in automated systems. The input data for the Expert Diagnostic System is a conflicting set of diagnostic codes generated by the automated system over the time interval ∆t during its operation. We proposed mathematical methods to work with a data structure “m-tuples based on ordinary sets of arbitrary cardinality n” to process the input data. We conducted a comparative analysis to estimate the execution time of algorithms for the diagnosis search formation space using sequential access to the Boolean of input data and using the method developed. The analysis showed that the application of the proposed method changes the functional dependency of the execution time estimation of the algorithm in accordance with the number of its input data n from exponential to cubic. The application of the method developed allows us to minimize the time needed to establish the diagnosis to real time. The method presented to diagnose automated systems allows creating methods and algorithms for automatic self-recovery of their operability after reversible failures in real time

Strength failure behavior of granite containing two holes under Brazilian test

A series of Brazilian tests under diameter compression for disc specimens was carried out to investigate the strength and failure behavior by using acoustic emission (AE) and photography monitoring technique. On the basis of experimental results, load-displacement curves, AE counts, real-time crack evolution process, failure modes and strength property of granite specimens containing two pre-existing holes were analyzed in detail. Two typical types of load-displacement curves are identified, i.e., sudden instability (type I) and progressive failure (type II). In accordance with the two types of load-displacement curves, the AE events also have different responses. The present experiments on disc specimens containing two pre-existing holes under Brazilian test reveal four distinct failure modes, including diametrical splitting failure mode (mode I), one crack coalescence failure mode (mode II), two crack coalescences failure mode (mode III) and no crack coalescence failure mode (mode IV). Compared with intact granite specimen, the disc specimen containing two holes fails with lower strength, which is closely related to the bridge angle. The failure strength of pre-holed specimen first decreases and then increases with the bridge angle. Finally, a preliminary interpretation was proposed to explain the strength evolution law of granite specimen containing two holes based on the microscopic observation of fracture plane.