Data-driven Condition Research Articles

A framework for advanced visualization of predictive analytics in cyber-physical production systems

This study discusses a framework to support manufacturers in their maintenance activities via data-driven condition monitoring and advanced visualization techniques. Moreover, the framework proposes a user-friendly predictive maintenance approach by enabling the user to configure the analysis chains by himself. Apart from this approach, the framework is built on containerization technologies to enable the platform to be flexible, resilient, and scalable in a wide range of production environments. The proposed techniques have been tested in a white goods use case with the preliminary results being reported in this work.

Simulation-Driven Deep Learning Approach for Wear Diagnostics in Hydrodynamic Journal Bearings

Abstract Early diagnosis in rotating machinery has been a challenge when looking toward the concept of intelligent machines. A crucial and critical component in these systems is the lubricated journal bearing, subjected to wear fault by abrasive removing of material in its inner wall, mainly during run-ups and run-downs. In extreme conditions, wear faults can cause unexpected shutdowns in rotating systems. Consequently, advanced condition monitoring is an essential procedure in the wear diagnosis of journal bearings. Although an increasing number of data-driven condition monitoring approaches for rotating machines have been proposed in the past decade, they mostly rely on substantial amounts of experimental data for training, which is expensive and time-consuming to obtain. The objective of this work is to develop a framework using a deep learning algorithm to classify wear faults in hydrodynamic journal bearings using simulated vibrations signals. Numerically simulated data sets under different wear severity levels and operating conditions were used to train and test the diagnostics framework. The results show that the proposed framework can be a promising tool to diagnose wear faults in journal bearings.

Domain Adaptation Digital Twin for Rolling Element Bearing Prognostics

Artificial Intelligence (AI) is escalating in data-driven condition monitoring research. Traditional expert knowledge-based Prognostics and Health Management (PHM) processes can be smartened up with the assistance of various AI techniques, such as deep learning models. On the other hand, current deep learning based prognostics suffers from the data deficit issue, especially considering the varying operating conditions and the degradation modes of the components in practical industrial applications. With the development of simulation techniques, physical-knowledge based digital twin models give engineers access to a large amount of simulation data at a lower cost. These simulation data contain the physical characteristics and the degradation information of the component. In order to accurately predict the Remaining Useful Life (RUL) during the degradation process, in this paper, a bearing digital twin model is constructed based on a phenomenological vibration model. A Domain Adversarial Neural Network (DANN) is used to achieve the domain adaptation target between the simulation and the real data. Regarding the simulation data as the source domain and real data as the target domain, the DANN model is able to predict the RUL without any priori knowledge of the labelling information. Based on real bearing run-to-failure experiments, the performance of the proposed method is validated with high RUL prediction accuracy.

Deep Learning-Based Domain Adaptation Method for Fault Diagnosis in Semiconductor Manufacturing

Quality inspection in semiconductor manufacturing is of great importance in the modern industries. In the recent years, intelligent data-driven condition monitoring methods have been successfully developed and applied in the industrial applications. However, despite the promising condition monitoring performance, the existing methods generally assume the training and testing data are from the same distribution. In practice, due to variations in manufacturing process, the collected data are usually subject to different distributions in different operating conditions, that significantly deteriorates the performance of the data-driven methods. To address this issue, this paper proposes a deep learning-based domain adaptation method for fault diagnosis in semiconductor manufacturing. The maximum mean discrepancy metric is optimized on the learned high-level data representation in the deep neural network. Experimental results on a real-world semiconductor manufacturing dataset suggest the proposed method offers an effective and generalized data-driven fault diagnosis approach for quality inspection.

A Data-Driven Condition Monitoring of Product Quality Analysis System Based on RS and AHP

A Data-Driven Condition Monitoring of Product Quality Analysis System Based on RS and AHP

Optimizing Maintenance Policies for a Yaw System Using Reliability-Centered Maintenance and Data-Driven Condition Monitoring

System downtime and unplanned outages massively affect plant productivity; therefore, the reliability, availability, maintainability, and safety (RAMS) disciplines, together with fault diagnosis and condition monitoring (CM), are mandatory in energy applications. This article focuses on the optimization of a maintenance plan for the yaw system used in an onshore wind turbine (WT). A complete reliability-centered maintenance (RCM) procedure is applied to the system to identify which maintenance action is the optimal solution in terms of cost, safety, and availability. The scope of the research is to propose a new customized decision-making diagram inside the RCM assessment to reduce the subjectivity of the procedure proposed in the standard and save the cost by optimizing maintenance decisions, making the projects more cost-efficient and cost-effective. This article concludes by proposing a new diagnostic method based on a data-driven CM system to efficiently monitor the health and detect damages in the WT by means of measurements of critical parameters of the tested system. This article highlights how a reliability analysis, during the early phase of the design, is a very helpful and powerful means to guide the maintenance decision and the data-driven CM.

Bridge condition rating data modeling using deep learning algorithm

This paper presents a deep learning-based bridge condition rating data modeling approach using selected data from the National Bridge Inventory (NBI) database. The objective of this research is to develop a data-driven approach that enables prediction of future conditions of highway bridge components from historical inspection data. The problem is solved by training a Convolutional Neural Network (CNN) model with online available NBI data. One prominent feature of the CNN model is that if well-trained it can represent the high dimensional data in the dataset abstractions for which conventional mathematical models may be difficult to describe. A case study of Maryland and Delaware highway bridges using historical data (1992–2017) sourced from the NBI database has been performed to demonstrate the proposed method. CNN models for three primary components of these highway bridges including the deck, superstructure, and substructure have been established. Optimization of model parameters is achieved through a parametric study. Research findings suggest that the deep learning model offers a promising tool as a data-driven condition forecasting approach for bridge components with a demonstrated prediction accuracy over 85%.

A Composite Anomaly Detection System for Data-Driven Power Plant Condition Monitoring

Data-driven condition monitoring is an essential function for power plant because of its potential to enhance asset longevity and reduce the operation and maintenance costs. This article explains the complicated relationship in multiplex power plant data as a mixture of temporal dependency and cross-variable association and proposes a composite anomaly detection system that incorporates the two data relationships on a probabilistic basis for more reliable power plant condition monitoring. It is able to dynamically capture the most significant relationship to develop more reliable normal condition interval, based on which the potential faults can be timely detected and the abnormal variable can be accurately identified. The proposed system was tested on a realistic thermal power plant. The testing results demonstrate its reliable condition monitoring and accurate anomaly detection performance, which necessitates the composite modeling of temporal dependency and cross-variable association in data-driven power plant condition monitoring.

Data-Driven Condition Monitoring Approaches to Improving Power Output of Wind Turbines

This paper presents data-driven approaches to improve the active power output of wind turbines based on estimating their health condition. The main procedure includes estimations of fault degree and health condition level, and optimal power dispatch control. The proposed method can adjust the active power output of individual turbines according to their health condition and can thus optimize the total energy output of a wind farm. In the paper, extreme learning machine algorithm and Bonferroni interval are applied to estimate fault degree, while an analytic hierarchy process is used to estimate the health condition level. A scheme for power dispatch control is formulated based on the estimated health condition. Models have been identified from supervisory control and data acquisition data acquired from an operational wind farm, which contains temperature data of gearbox bearing and generator winding. The results show that the proposed method can maximize the operation efficiency of the wind farm while significantly reducing the fatigue loading on the faulty wind turbines.

Alarms‐related wind turbine fault detection based on kernel support vector machines

Wind power is playing an increasingly significant role in daily life. However, wind farms are usually far away from cities especially for offshore wind farms, which brought inconvenience for maintenance. Two conventional maintenance strategies, namely corrective maintenance and preventive maintenance, cannot provide condition-based maintenance to identify potential anomalies and predicts turbines' future operation trend. In this study, a model based data-driven condition monitoring method is proposed for fault detection of the wind turbines (WTs) with SCADA data acquired from an operational wind farm. Due to the nature of the alarm signals, the alarm data can be used as an intermedium to link the normal data and fault data. First, KPCA is employed to select principal components (PCs) to retain the dominant information from the original dataset to reduce the computation load for further modelling. Then the selected PCs are processed for normal-abnormal condition classification to extract those abnormal condition data that are classified further into false alarms and true alarms related to the faults. This two-stage classification approach is implemented based on the KSVM algorithm. The results demonstrate that the two-stage fault detection method can identify the normal, alarm and fault conditions of the WTs accurately and effectively.

Monitoring Wind Turbine Gearbox with Echo State Network Modeling and Dynamic Threshold Using SCADA Vibration Data

Health monitoring of wind turbine gearboxes has gained considerable attention as wind turbines become larger in size and move to more inaccessible locations. To improve the reliability, extend the lifetime of the turbines, and reduce the operation and maintenance cost caused by the gearbox faults, data-driven condition motoring techniques have been widely investigated, where various sensor monitoring data (such as power, temperature, and pressure, etc.) have been modeled and analyzed. However, wind turbines often work in complex and dynamic operating conditions, such as variable speeds and loads, thus the traditional static monitoring method relying on a certain fixed threshold will lead to unsatisfactory monitoring performance, typically high false alarms and missed detections. To address this issue, this paper proposes a reliable monitoring model for wind turbine gearboxes based on echo state network (ESN) modeling and the dynamic threshold scheme, with a focus on supervisory control and data acquisition (SCADA) vibration data. The aim of the proposed approach is to build the turbine normal behavior model only using normal SCADA vibration data, and then to analyze the unseen SCADA vibration data to detect potential faults based on the model residual evaluation and the dynamic threshold setting. To better capture temporal information inherent in monitored sensor data, the echo state network (ESN) is used to model the complex vibration data due to its simple and fast training ability and powerful learning capability. Additionally, a dynamic threshold monitoring scheme with a sliding window technique is designed to determine dynamic control limits to address the issue of the low detection accuracy and poor adaptability caused by the traditional static monitoring methods. The effectiveness of the proposed monitoring method is verified using the collected SCADA vibration data from a wind farm located at Inner Mongolia in China. The results demonstrated that the proposed method can achieve improved detection accuracy and reliability compared with the traditional static threshold monitoring method.

Forward-Looking Element Recognition Based on the LSTM-CRF Model with the Integrity Algorithm

A state-of-the-art entity recognition system relies on deep learning under data-driven conditions. In this paper, we combine deep learning with linguistic features and propose the long short-term memory-conditional random field model (LSTM-CRF model) with the integrity algorithm. This approach is primarily based on the use of part-of-speech (POS) syntactic rules to correct the boundaries of LSTM-CRF model annotations and improve its performance by raising the integrity of the elements. The method incorporates the advantages of the data-driven method and dependency syntax, and improves the precision rate of the elements without losing recall rate. Experiments show that the integrity algorithm is not only easy to combine with the other neural network model, but the overall effect is better than several advanced methods. In addition, we conducted cross-domain experiments based on a multi-industry corpus in the financial field. The results indicate that the method can be applied to other industries.

Issues with Data Quality for Wind Turbine Condition Monitoring and Reliability Analyses

In order to remain competitive, wind turbines must be reliable machines with efficient and effective maintenance strategies. However, thus far, wind turbine reliability information has been closely guarded by the original equipment manufacturers (OEMs), and turbine reliability studies often rely on data that are not always in a usable or consistent format. In addition, issues with turbine maintenance logs and alarm system data can make it hard to identify historical periods of faulty operation. This means that building new and effective data-driven condition monitoring techniques and methods can be challenging, especially those that rely on supervisory control and data acquisition (SCADA) system data. Such data are rarely standardised, resulting in challenges for researchers in contextualising these data. This work aims to summarise some of the issues seen in previous studies, highlighting the common problems seen by researchers working in the areas of condition monitoring and reliability analysis. Standards and policy initiatives that aim to alleviate some of these problems are given, and a summary of their recommendations is presented. The main finding from this work is that industry would benefit hugely from unified standards for turbine taxonomies, alarm codes, SCADA operational data and maintenance and fault reporting.

Monitoring, Modeling, and Assessment of a Self-Sensing Railway Bridge during Construction

This study shows how integrating fiber optic sensor (FOS) networks into bridges during the construction stage can be used to quantify preservice performance. Details of the installation of a large FOS network on a new steel-concrete composite railway bridge in the United Kingdom are presented. An overview of the FOS technology, installation techniques, and monitoring program is also presented, and the monitoring results from several construction stages are discussed. A finite-element (FE) model was developed and a phased analysis was carried out to simulate strain development in the bridge during consecutive construction stages. The response of the self-sensing bridge to the time-dependent properties of the concrete deck was evaluated by comparing FOS measurements to predicted results according to several model code formulations implemented in the FE model. The preservice strain distribution due to dead loading is typically assumed to act uniformly along the bridge length; however, the monitoring results revealed that the distribution was highly variable as a result of the complex interactions between gravity loading, bridge geometry, time-dependent concrete properties, and temperature effects. Moment utilization of the main girders and composite beams, during preservice conditions, was assessed and found to be between 19.3 and 24.9% of the respective design cross-section capacities. Quantifying preservice performance via integrated sensing also provided a critical baseline for the bridge, which enables future data-driven condition assessments.

Data-driven condition now- and forecasting of railway switches for

This contribution presents a data-based model that exploitsthe power consumed by point engines during bladesmovement of railway switches to detect relevant anomaliesin switch behavior. The model incorporates local airtemperature at the time of the measurement to account forthe significant influence of the environmental conditions onnormal switch behavior. Anomaly detection by the model isvalidated against alerts triggered by the state-of-the-artmonitoring system POSS®, which is based on switchspecific and manually selected reference curves. The databased model leads to less in number and more reliable alertsin comparison to the current version of POSS®. Especiallyfalse alerts caused by temperature effects are significantlyreduced. Furthermore, the high sensitivity of the modelproves to be capable of detecting emerging switch failures atan early stage of development. The detection capabilities ofswitch condition (nowcast) and identification of emergingfailures at an early stage (required for failure forecast)proves that the model is useful for traffic interferenceprevention, condition-based predictive maintenance andswitch health enhancement.

A graphic pattern feature-mapping-based data-driven condition monitoring method for diesel engine malfunction identification and classification

Effective condition monitoring of diesel engine can ensure the reliability of large-power machines and prevent catastrophic consequences. Cylinder pressure is capable of reflecting the whole combustion process of diesel engine, and hence can help to identify the malfunctions of the diesel engine during operation. In this paper, a graphic pattern feature-mapping method is proposed for graphic pattern feature recognition in data-driven condition monitoring. The graphic feature extraction and recognition are linked by labeled feature-mapping. It is used for identifying the running condition of the diesel engine via analyzing the cylinder pressure signal of the diesel engine. The different types of the malfunctions which are caused by different parts of the diesel engine such as induction system, valve actuating mechanism, fuel system, fuel injection system, etc. can be identified just by cylinder pressure signal. The bench experiment of a large-power diesel engine is performed to validate this graphic pattern recognition method. The results show that it has good accuracy on multi-malfunction identification and classification when the engine operates at one speed and one load.

Life Prediction of Bearing by Using Adaboost Regressor

Rolling element bearings are very important and highly utilized in many industries. Their catastrophic failure due to fluctuating working conditions leads to unscheduled breakdown and increases accidental economical losses. Thus these issues have triggered a need for reliable and automatic prognostics methodology which will prevent a potentially expensive maintenance program. Accordingly, Remaining Useful Life (RUL) prediction based on artificial intelligence is an attractive methodology for several researchers. In this study, data-driven condition monitoring approach is implemented for predicting RUL of bearing under a certain load and speed. The approach demonstrates the use of complex boosting ensemble like AdaBoost Regressor for prediction of RUL with time-domain features which are extracted by applying various statistical functions on given vibration signals. Decision based Feature Ranking Method is used to find the contribution of each feature. Hyper-parameters are tuned for these models by using exhaustive parameter search and performance of these models is further verified by plotting respective learning curves. In experimental analysis, the proposed approach is verified using FEMTO bearing data-set provided by IEEE PHM Data Challenge 2012. Weibull Hazard Rate Function for each bearing from learning data set are used to find target values i.e. projected RUL of the bearings. Ensemble models are trained using extracted time domain features and target values of learning data set which contains vibration signals of six run-to-failure bearings with three different combinations of speed and load. Vibration data set of eleven partially degraded bearings from same three groups are fed as testing data set to trained models to estimate their RULs. Appropriate scoring function is used to calculate score of models using error between given real RULs and predicted values of test set bearings. Testing scores of proposed models are compared with well-established data-driven approaches from literature and are found to be better than all the models applied on this data-set, thereby demonstrating the reliability of the proposed model.

Bearing Health Condition Prediction Using Deep Belief Network

Bearings play a critical role in maintaining safety and reliability of rotating machinery. Bearings health condition prediction aims to prevent unexpected failures and minimize overall maintenance costs since it provides decision making information for condition-based maintenance. This paper proposes a Deep Belief Network (DBN)-based data-driven health condition prediction method for bearings. In this prediction method, a DBN is used as the predictor, which includes stacked RBMs and regression output. Our main contributions include development of a deep leaning-based data-driven prognosis solution that does not rely on explicit model equations and prognostic expertise, and providing comprehensive prediction results on five representative run-to-failure bearings. The IEEE PHM 2012 challenge dataset is used to demonstrate the effectiveness of the proposed method, and the results are compared with two existing methods. The results show that the proposed method has promising performance in terms of short-term health condition prediction and remaining useful life prediction for bearings.

Improved Railway Track Irregularities Classification by aModel Inversion Approach

Over time railway networks have become complex systems characterized by manifold types of technical components with a broad range of age distribution. De facto, about 50 percent of the life cycle costs of railway infrastructures aremade up by direct and indirect maintenance costs. A remedy can be provided by a condition based preventive maintenance strategy leading to an optimized scheduling of maintenance actions taking the actual aswell as the expected future infrastructure condition into account. A prerequisite is, however, that the thousands of kilometers of railway tracks are almost continuously monitored. Thus, a promising approach is the usage of low-cost sensors, e.g. accelerometers and gyroscopes, which can be installed on common in-line freight and passenger trains. Due to ambiguous data records a credible classification of railway track irregularities directly from these data is challenging. Alternatively to this pure data-driven approach, in this paper a novel hybrid approach is presented. To this end, a simplified vehicle suspension model is applied for the purpose of railway track condition monitoring by analyzing the dynamic railway track - train interactions. The inversion of the model can be used to recalculate the actual inputs (irregularities) of the monitored system (rail surface) which have caused recorded system responses (dynamic vehicle reactions and acceleration data, respectively). These recalculated inputs are a sound basis of subsequent data-driven condition monitoring analyses.In this preliminary study, a classification algorithm is implemented to identify a simulated railway track irregularity automatically.

More data-driven processing at retrieval reduces age-related memory deficits.

The environmental support hypothesis postulates that it may be possible to reduce older adults' deficits in episodic memory by providing environmental support at the encoding and/or retrieval phases. To examine the validity of this hypothesis, we varied the amount of retrieval support by manipulating data-driven processes. Young and older adults performed a word-stem cued recall task under a low data-driven condition (LDDC) in which the retrieval cue comprised 3 letters, and a higher data-driven condition (HDDC) in which the cue comprised 4 letters. Older adults benefitted more than younger adults from the additional support. Older adults exhibited a large deficit relative to younger adults in the LDDC condition but no age differences were found in the HDDC condition. These findings demonstrate that age-related memory deficits can be reduced by increasing the environmental support at retrieval associated with the data-driven component of retrieval processing.