Pub Date : 2022-06-29DOI: 10.36001/phme.2022.v7i1.3340
M. Rahat, P. Mashhadi, Sławomir Nowaczyk, T. Rognvaldsson, Atabak Taheri, A. Abbasi
The discrepancy in the distribution of source and target domains is usually referred to as a domain shift. It is one of the reasons for the inferior performance of machine learning solutions at deployment. We illustrate that the domain shift issue is pertinent to the readings of the vehicles’ operational sensors. This is due to the fact that these measurements are collected over a period of time and are susceptible to various changes that happen in the meantime. Examples of these changes are usage pattern variations, aging of the vehicles, seasonal shifts, and driver changes. However, domain adversarial neural networks (DANN) have shown promising results to reduce the negative impact of the domain shift. The present study investigates domain adaptation (DA) in the predictive maintenance field by estimating the remaining useful life (RUL) of turbochargers. The devices are operating on a fleet of VOLVO trucks, and the information about their services is collected over four years between 2016 and 2019. The input features to the model are a set of bi-weekly collected measurements called logged vehicle data (LVD). The contributions of this paper are two-fold. First, we propose a new approach for detecting domain (covariate) shift using an autoencoder. Second, we adapt domain adversarial neural networks to the specific application of predicting turbocharger failures. Finally, we deploy a recurrent feature extraction layer in the DANN architecture to incorporate temporal aspect of the data. The experimental results demonstrate the superiority of the proposed method over the traditional approach.
{"title":"Domain Adaptation in Predicting Turbocharger Failures Using Vehicle’s Sensor Measurements","authors":"M. Rahat, P. Mashhadi, Sławomir Nowaczyk, T. Rognvaldsson, Atabak Taheri, A. Abbasi","doi":"10.36001/phme.2022.v7i1.3340","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.3340","url":null,"abstract":"The discrepancy in the distribution of source and target domains is usually referred to as a domain shift. It is one of the reasons for the inferior performance of machine learning solutions at deployment. We illustrate that the domain shift issue is pertinent to the readings of the vehicles’ operational sensors. This is due to the fact that these measurements are collected over a period of time and are susceptible to various changes that happen in the meantime. Examples of these changes are usage pattern variations, aging of the vehicles, seasonal shifts, and driver changes. However, domain adversarial neural networks (DANN) have shown promising results to reduce the negative impact of the domain shift. The present study investigates domain adaptation (DA) in the predictive maintenance field by estimating the remaining useful life (RUL) of turbochargers. The devices are operating on a fleet of VOLVO trucks, and the information about their services is collected over four years between 2016 and 2019. The input features to the model are a set of bi-weekly collected measurements called logged vehicle data (LVD). The contributions of this paper are two-fold. First, we propose a new approach for detecting domain (covariate) shift using an autoencoder. Second, we adapt domain adversarial neural networks to the specific application of predicting turbocharger failures. Finally, we deploy a recurrent feature extraction layer in the DANN architecture to incorporate temporal aspect of the data. The experimental results demonstrate the superiority of the proposed method over the traditional approach.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123511859","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-29DOI: 10.36001/phme.2022.v7i1.3364
Alon Sol, E. Madar, J. Bortman, R. Klein
To date, much of the research done in the field of condition monitoring of rotating machinery is conducted in the frequency domain. The frequency domain analysis specifically for bearings is based on extracting features from the spectrum of the vibration signature. These features are mostly based on the amplitude at the bearing tones along with their sidebands and high order harmonics. Therefore, it is important to determine the location of the mentioned bearing tones in the spectrum accurately and automatically. For the case of ball bearings this process can be problematic due to slippage of the rolling elements and variations in the angle of contact. These may cause the bearing tone to deviate from its nominal value. To this day, the common practice for bearing diagnostics is based on the vibration level at the analytical bearing tones or involvement of experts to identify the true location of the bearing tone. In this research an autonomous algorithm for bearing tone extraction, based on pattern matching, was developed. The proposed algorithm is based on the common assumption that the spectrum of a faulted bearing contains a certain known pattern of prominent peaks. The algorithm “scans” the entire spectrum and determines the frequency value which has the highest correlation to the mentioned pattern. The proposed algorithm was validated and its capabilities are illustrated using experimental data. This algorithm is able to assist any diagnostic approach towards automatic and reliable feature extraction process, both for physics based and data driven approaches.
{"title":"Autonomous Bearing Tone Tracking Algorithm","authors":"Alon Sol, E. Madar, J. Bortman, R. Klein","doi":"10.36001/phme.2022.v7i1.3364","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.3364","url":null,"abstract":"To date, much of the research done in the field of condition monitoring of rotating machinery is conducted in the frequency domain. The frequency domain analysis specifically for bearings is based on extracting features from the spectrum of the vibration signature. These features are mostly based on the amplitude at the bearing tones along with their sidebands and high order harmonics. Therefore, it is important to determine the location of the mentioned bearing tones in the spectrum accurately and automatically. For the case of ball bearings this process can be problematic due to slippage of the rolling elements and variations in the angle of contact. These may cause the bearing tone to deviate from its nominal value. \u0000To this day, the common practice for bearing diagnostics is based on the vibration level at the analytical bearing tones or involvement of experts to identify the true location of the bearing tone. In this research an autonomous algorithm for bearing tone extraction, based on pattern matching, was developed. The proposed algorithm is based on the common assumption that the spectrum of a faulted bearing contains a certain known pattern of prominent peaks. The algorithm “scans” the entire spectrum and determines the frequency value which has the highest correlation to the mentioned pattern. \u0000The proposed algorithm was validated and its capabilities are illustrated using experimental data. This algorithm is able to assist any diagnostic approach towards automatic and reliable feature extraction process, both for physics based and data driven approaches.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125507201","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-29DOI: 10.36001/phme.2022.v7i1.2784
Alaaeddine Chaoub, Christophe Cerisara, A. Voisin, B. Iung
Deep learning (DL) approaches have multiple potential advantages that have been explored in various fields, but for prognostic and health management (PHM) applications, this is not the case due to the lack of data in particular applications and also due of the absence of multiple DL-oriented benchmarks as in other fields, which limits the research in this area even though these types of applications will have a strong impact on the industrial world. To introduce the benefits of DL in this area, we should be able to develop models even when we have small data sets, to verify whether or not this is possible, in this thesis we explore the research direction of few shot learning in the context of equipment PHM.
{"title":"Deep Learning Representation Pre-training for Industry 4.0","authors":"Alaaeddine Chaoub, Christophe Cerisara, A. Voisin, B. Iung","doi":"10.36001/phme.2022.v7i1.2784","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.2784","url":null,"abstract":"Deep learning (DL) approaches have multiple potential advantages that have been explored in various fields, but for prognostic and health management (PHM) applications, this is not the case due to the lack of data in particular applications and also due of the absence of multiple DL-oriented benchmarks as in other fields, which limits the research in this area even though these types of applications will have a strong impact on the industrial world. To introduce the benefits of DL in this area, we should be able to develop models even when we have small data sets, to verify whether or not this is possible, in this thesis we explore the research direction of few shot learning in the context of equipment PHM.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121676704","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-29DOI: 10.36001/phme.2022.v7i1.3366
Sylvain Poupry, Cédrick Béler, K. Medjaher
Today, air quality monitoring is a global concern. The World Health Organization (WHO) defined standards for each pollutant and each member state is committed to monitoring them continuously and reliably to protect the population. This responsibility is delegated to air quality monitoring associations. To achieve the objectives of reliable, accurate, and continuous measurements, these associations rely on conventional measuring stations with demanding specifications to serve as scientific references and decision supports for the authorities. However, because of heavy investments and required qualified staff, there are few stations and the coverage is coarse for territories of several thousand km2. To circumvent this difficulty, measurement network architectures using Low-Cost Sensors (LCS) have been deployed. Low cost and requiring less qualification, This alternative technology to conventional measuring stations makes it possible to target local pollution that could not otherwise be detected. Although it is more accurate on the spatial dimension, this technology has several drawbacks, notably in terms of measurement repeatability and hardware quality. It is also subject to measurement drifts over time. To overcome these drawbacks, a resilient and reliable architecture based on LCS and triple redundancy has been proposed. The basic principle is based on the implementation of three smart sensors (SmS) using LCS measuring the same parameters on the same perimeter. These SmS communicate with an Aggregator that aggregates the data sent by SmS. The aggregator includes also detection and voting tasks allowing to compare, cross the data, detect faults of LCS online, and provide data that are ready for processing. In this paper, a pre-processing algorithm in four steps is presented. It identifies hardware faults from one or more LCS and reports outliers for verification by an expert. It is configurable and can identify failure behaviors (LCS or air quality). Finally, the proposed algorithm excludes the outliers data from faulty LCS and presents only reliable ones.
{"title":"Towards Data Reliability Based on Triple Redundancy and Online Outlier Detection","authors":"Sylvain Poupry, Cédrick Béler, K. Medjaher","doi":"10.36001/phme.2022.v7i1.3366","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.3366","url":null,"abstract":"Today, air quality monitoring is a global concern. The World Health Organization (WHO) defined standards for each pollutant and each member state is committed to monitoring them continuously and reliably to protect the population. This responsibility is delegated to air quality monitoring associations. To achieve the objectives of reliable, accurate, and continuous measurements, these associations rely on conventional measuring stations with demanding specifications to serve as scientific references and decision supports for the authorities. However, because of heavy investments and required qualified staff, there are few stations and the coverage is coarse for territories of several thousand km2. To circumvent this difficulty, measurement network architectures using Low-Cost Sensors (LCS) have been deployed. Low cost and requiring less qualification, This alternative technology to conventional measuring stations makes it possible to target local pollution that could not otherwise be detected. Although it is more accurate on the spatial dimension, this technology has several drawbacks, notably in terms of measurement repeatability and hardware quality. It is also subject to measurement drifts over time. To overcome these drawbacks, a resilient and reliable architecture based on LCS and triple redundancy has been proposed. The basic principle is based on the implementation of three smart sensors (SmS) using LCS measuring the same parameters on the same perimeter. These SmS communicate with an Aggregator that aggregates the data sent by SmS. The aggregator includes also detection and voting tasks allowing to compare, cross the data, detect faults of LCS online, and provide data that are ready for processing. In this paper, a pre-processing algorithm in four steps is presented. It identifies hardware faults from one or more LCS and reports outliers for verification by an expert. It is configurable and can identify failure behaviors (LCS or air quality). Finally, the proposed algorithm excludes the outliers data from faulty LCS and presents only reliable ones.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125746989","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-29DOI: 10.36001/phme.2022.v7i1.3365
Weikun Deng, K. Nguyen, C. Gogu, J. Morio, K. Medjaher
This paper proposes hybrid methods using physics-informed (PI) lightweight Temporal Convolution Neural Network (PITCN) for bearings’ remaining useful life (RUL) prediction under stiffness degradation. It includes three PI hybrid models: a) PI Feature model (PIFM) — constructing physics-informed health indicator (PIHI) to augment the feature space, b) PI Layer model (PILM)—encoding the physics governing equations in a hidden layer, and c) PI Layer Based Loss model (PILLM)—designing PI conflict loss, taking into account the difference before and after integration of the physics input-output relations involved module to the loss function. We simulated 200 different bearing stiffness degradations, using their discrete monitored vibration signals to verify the effectiveness of the proposed method. We also investigate their inference process through feature heat map analysis to interpret how the models melt physics knowledge to assist in capturing the degradation trend. The physics knowledge considered in this paper is the dynamic relationship between vibration amplitude and stiffness in a damped forced vibration model. The results show that all three PITCN models effectively capture degradation-related trend information and perform better than the vanilla lightweight TCN. Furthermore, the visualization of the feature channels highlights the important role of physics information in model training. Channels containing physics information demonstrate higher correlation with results as they significantly dominate the heat map compared to other channels.
{"title":"Physics-informed Lightweight Temporal Convolution Networks for Fault Prognostics Associated to Bearing Stiffness Degradation","authors":"Weikun Deng, K. Nguyen, C. Gogu, J. Morio, K. Medjaher","doi":"10.36001/phme.2022.v7i1.3365","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.3365","url":null,"abstract":"This paper proposes hybrid methods using physics-informed (PI) lightweight Temporal Convolution Neural Network (PITCN) for bearings’ remaining useful life (RUL) prediction under stiffness degradation. It includes three PI hybrid models: a) PI Feature model (PIFM) — constructing physics-informed health indicator (PIHI) to augment the feature space, b) PI Layer model (PILM)—encoding the physics governing equations in a hidden layer, and c) PI Layer Based Loss model (PILLM)—designing PI conflict loss, taking into account the difference before and after integration of the physics input-output relations involved module to the loss function. We simulated 200 different bearing stiffness degradations, using their discrete monitored vibration signals to verify the effectiveness of the proposed method. We also investigate their inference process through feature heat map analysis to interpret how the models melt physics knowledge to assist in capturing the degradation trend. The physics knowledge considered in this paper is the dynamic relationship between vibration amplitude and stiffness in a damped forced vibration model. The results show that all three PITCN models effectively capture degradation-related trend information and perform better than the vanilla lightweight TCN. Furthermore, the visualization of the feature channels highlights the important role of physics information in model training. Channels containing physics information demonstrate higher correlation with results as they significantly dominate the heat map compared to other channels.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126689815","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper focuses on the problem of predicting production line status for Printed Circuit Boards (PCBs). The problem contains three prediction tasks regarding PCB producing process. Firstly, data exploration is carried out and it reveals several data challenges, including data imbalance, data noise, small sample size, and component difference. To predict production line status for components of PCBs using records of inspection on pins, we proposed two possible feature extraction methods to compress the pin-level data into component-level. A statistical feature extraction method, which retrieves descriptive statistics such as mean, standard deviation, maximum, and minimum of pins on the same component, is applied to Task 1, while a PinNumber-based feature extraction method, which keep original values for 2-pin components, is applied to Task3. In addition, a neural-net model with feeding imbalance control is established for Task 1. and a random forests model is applied for both Task 2 and Task 3. Moreover, a threshold moving technique is proposed to optimize the threshold selection. Finally, the result shows that our models achieved f1-scores of 0.44, 0.54, and 0.71 on the test set for the three tasks, respectively.
{"title":"Prediction of Production Line Status for Printed Circuit Boards","authors":"Haichuan Tang, Yin Tian, Junyan Dai, Yuan Wang, Jian-li Cong, Qi Liu, Xuejun Zhao, Yunxiao Fu","doi":"10.36001/phme.2022.v7i1.3371","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.3371","url":null,"abstract":"This paper focuses on the problem of predicting production line status for Printed Circuit Boards (PCBs). The problem contains three prediction tasks regarding PCB producing process. Firstly, data exploration is carried out and it reveals several data challenges, including data imbalance, data noise, small sample size, and component difference. To predict production line status for components of PCBs using records of inspection on pins, we proposed two possible feature extraction methods to compress the pin-level data into component-level. A statistical feature extraction method, which retrieves descriptive statistics such as mean, standard deviation, maximum, and minimum of pins on the same component, is applied to Task 1, while a PinNumber-based feature extraction method, which keep original values for 2-pin components, is applied to Task3. In addition, a neural-net model with feeding imbalance control is established for Task 1. and a random forests model is applied for both Task 2 and Task 3. Moreover, a threshold moving technique is proposed to optimize the threshold selection. Finally, the result shows that our models achieved f1-scores of 0.44, 0.54, and 0.71 on the test set for the three tasks, respectively.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125129012","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-29DOI: 10.36001/phme.2022.v7i1.3320
Ingeborg de Pater, M. Mitici
Well-established metrics such as the Root Mean Square Error or the Mean Absolute Error are not suitable to evaluate estimated distributions of the Remaining Useful Life (i.e., probabilistic prognostics). We therefore propose novel metrics to evaluate the quality of probabilistic Remaining Useful Life prognostics. We estimate the distribution of the Remaining Useful Life of turbofan engines using a Convolutional Neural Network with Monte Carlo dropout. The accuracy and sharpness of the obtained probabilistic prognostics are evaluated using the Continuous Ranked Probability Score (CRPS) and weighted CRPS. The reliability of the obtained probabilistic prognostics is evaluated using the α-Coverage and the Reliability Score. The results show that the estimated distributions of the Remaining Useful Life of turbofan engines are accurate, reliable and sharp when using a Convolutional Neural Network with Monte Carlo dropout. In general, the proposed metrics are suitable to evaluate the accuracy, sharpness and reliability of probabilistic Remaining Useful Life prognostics.
{"title":"Novel Metrics to Evaluate Probabilistic Remaining Useful Life Prognostics with Applications to Turbofan Engines","authors":"Ingeborg de Pater, M. Mitici","doi":"10.36001/phme.2022.v7i1.3320","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.3320","url":null,"abstract":"Well-established metrics such as the Root Mean Square Error or the Mean Absolute Error are not suitable to evaluate estimated distributions of the Remaining Useful Life (i.e., probabilistic prognostics). We therefore propose novel metrics to evaluate the quality of probabilistic Remaining Useful Life prognostics. We estimate the distribution of the Remaining Useful Life of turbofan engines using a Convolutional Neural Network with Monte Carlo dropout. The accuracy and sharpness of the obtained probabilistic prognostics are evaluated using the Continuous Ranked Probability Score (CRPS) and weighted CRPS. The reliability of the obtained probabilistic prognostics is evaluated using the α-Coverage and the Reliability Score. The results show that the estimated distributions of the Remaining Useful Life of turbofan engines are accurate, reliable and sharp when using a Convolutional Neural Network with Monte Carlo dropout. In general, the proposed metrics are suitable to evaluate the accuracy, sharpness and reliability of probabilistic Remaining Useful Life prognostics.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131202412","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-29DOI: 10.36001/phme.2022.v7i1.3321
Jennifer Blair, B. Stephen, Blair Brown, Alistair Forbes, S. Mcarthur
Nuclear plant operators are required to understand the uncertainties associated with the deployment of prognostics toolsin order to justify their inclusion in operational decision making processes and satisfy regulatory requirements. Operationaluncertainty can cause underlying prognostics models to underperform on assets that are subject to evolving impactsof age, manufacturing tolerances, operating conditions, and operating environment effects, of which may be capturedthrough a condition monitoring (CM) system that itself may be degraded. Sources of uncertainty in the data acquisitionpipeline can impact the health of CM data used to estimate the remaining useful life (RUL) of assets. These uncertaintiescan disguise or misrepresent developing faults, where (for example) the fault identification is not achieved until it hasprogressed to an unmanageable state. This leaves little flexibility for the operator’s maintenance decisions and generallyundermines model confidence. One method to quantify and account for operational uncertainty is calibrated hybrid models, employing physics, knowledgeor data driven methods to improve model accuracy and robustness. Hybrid models allow known physical relations tooffset full reliance on potentially untrustworthy data, whilst reducing the need for an abundance of representative historicaldata to reliably identify the monitored asset’s underlying behavioural trends. Calibration of the model then ensuresthe model is updated and representative of the real monitored asset by accounting for differences between the physics orknowledge model and CM data. In this paper, an open-source bearing knowledge informed machine learning (ML) model and CM datasets are utilizedin an illustrative bearing prognostic application. The uncertainty incurred by the decisions made at key stages in thedevelopment of the model’s data acquisition and processing pipeline are assessed and demonstrated by the resultant impacton RUL prediction performance. It was shown that design decisions could result in multiple valid pipeline designswhich generated different predicted RUL trajectories, increasing the uncertainty in the model output.
{"title":"Hybrid Fault Prognostics for Nuclear Applications: Addressing Rotating Plant Model Uncertainty","authors":"Jennifer Blair, B. Stephen, Blair Brown, Alistair Forbes, S. Mcarthur","doi":"10.36001/phme.2022.v7i1.3321","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.3321","url":null,"abstract":"Nuclear plant operators are required to understand the uncertainties associated with the deployment of prognostics toolsin order to justify their inclusion in operational decision making processes and satisfy regulatory requirements. Operationaluncertainty can cause underlying prognostics models to underperform on assets that are subject to evolving impactsof age, manufacturing tolerances, operating conditions, and operating environment effects, of which may be capturedthrough a condition monitoring (CM) system that itself may be degraded. Sources of uncertainty in the data acquisitionpipeline can impact the health of CM data used to estimate the remaining useful life (RUL) of assets. These uncertaintiescan disguise or misrepresent developing faults, where (for example) the fault identification is not achieved until it hasprogressed to an unmanageable state. This leaves little flexibility for the operator’s maintenance decisions and generallyundermines model confidence. \u0000One method to quantify and account for operational uncertainty is calibrated hybrid models, employing physics, knowledgeor data driven methods to improve model accuracy and robustness. Hybrid models allow known physical relations tooffset full reliance on potentially untrustworthy data, whilst reducing the need for an abundance of representative historicaldata to reliably identify the monitored asset’s underlying behavioural trends. Calibration of the model then ensuresthe model is updated and representative of the real monitored asset by accounting for differences between the physics orknowledge model and CM data. \u0000In this paper, an open-source bearing knowledge informed machine learning (ML) model and CM datasets are utilizedin an illustrative bearing prognostic application. The uncertainty incurred by the decisions made at key stages in thedevelopment of the model’s data acquisition and processing pipeline are assessed and demonstrated by the resultant impacton RUL prediction performance. It was shown that design decisions could result in multiple valid pipeline designswhich generated different predicted RUL trajectories, increasing the uncertainty in the model output.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"530 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132503966","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-29DOI: 10.36001/phme.2022.v7i1.3367
M. Asres, G. Cummings, A. Khukhunaishvili, P. Parygin, S. Cooper, D. Yu, J. Dittmann, C. Omlin
Predictive maintenance is essential for complex industrial systems to foresee anomalies before major system faults or ultimate breakdown. However, the existing efforts on Industry 4.0 predictive monitoring are directed at semi-supervised anomaly detection with limited robustness for large systems, which are often accompanied by uncleaned and unlabeled data. We address the challenge of predicting anomalies through data-driven end-to-end deep learning models using early warning symptoms on multivariate time series sensor data. We introduce AnoP, a long multi-timestep anomaly prediction system based on unsupervised attention-based causal residual networks, to raise alerts for anomaly prevention. The experimental evaluation on large data sets from detector health monitoring of the Hadron Calorimeter of the CMS Experiment at LHC CERN demonstrates the promising efficacy of the proposed approach. AnoP predicted around 60% of the anomalies up to seven days ahead, and the majority of the missed anomalies are abnormalities with unpredictable noisy-like behavior. Moreover, it has discovered previously unknown anomalies in the calorimeter’s sensors.
{"title":"Long Horizon Anomaly Prediction in Multivariate Time Series with Causal Autoencoders","authors":"M. Asres, G. Cummings, A. Khukhunaishvili, P. Parygin, S. Cooper, D. Yu, J. Dittmann, C. Omlin","doi":"10.36001/phme.2022.v7i1.3367","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.3367","url":null,"abstract":"Predictive maintenance is essential for complex industrial systems to foresee anomalies before major system faults or ultimate breakdown. However, the existing efforts on Industry 4.0 predictive monitoring are directed at semi-supervised anomaly detection with limited robustness for large systems, which are often accompanied by uncleaned and unlabeled data. We address the challenge of predicting anomalies through data-driven end-to-end deep learning models using early warning symptoms on multivariate time series sensor data. We introduce AnoP, a long multi-timestep anomaly prediction system based on unsupervised attention-based causal residual networks, to raise alerts for anomaly prevention. The experimental evaluation on large data sets from detector health monitoring of the Hadron Calorimeter of the CMS Experiment at LHC CERN demonstrates the promising efficacy of the proposed approach. AnoP predicted around 60% of the anomalies up to seven days ahead, and the majority of the missed anomalies are abnormalities with unpredictable noisy-like behavior. Moreover, it has discovered previously unknown anomalies in the calorimeter’s sensors.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"212 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116428031","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-29DOI: 10.36001/phme.2022.v7i1.3347
L. Baldo, P. Berri, M. D. Dalla Vedova, P. Maggiore
The growing adoption of electrical energy as a secondary form of onboard power leads to an increase of electromechanical actuators (EMAs) use in aerospace applications. Therefore, innovative prognostic and diagnostic methodologies are becoming a fundamental tool to early identify faults propagation, prevent performance degradation, and ensure an acceptable level of safety and reliability of the system. Furthermore, prognostics entails further advantages, including a better ability to plan the maintenance of the various equipment, manage the warehouse and maintenance personnel, and a reduction in system management costs. Frequently, such approaches require the development of typologies of numerical models capable of simulating the performance of the EMA with different levels of fidelity: monitoring models, suitably simplified to combine speed and accuracy with reduced computational costs, and high fidelity models (and high computational intensity), to generate databases, develop predictive algorithms and train machine learning surrogates. Because of this, the authors developed a high-fidelity multi-domain numerical model (HF) capable of accounting for a variety of physical phenomena and gradual failures in the EMA, as well as a low-fidelity counterpart (LF). This simplified model is derived by the HF and intended for monitoring applications. While maintaining a low computing cost, LF is fault sensitive and can simulate the system position, speed, and equivalent phase currents. These models have been validated using a dedicated EMA test bench, designed and implemented by authors. The HF model can simulate the operation of the actuator in nominal conditions as well as in the presence of incipient mechanical faults, such as a variation in friction and an increase of backlash in the reduction gearbox. Comparing the preliminary results highlights satisfactory consistency between the experimental test bench and the two numerical models proposed by the authors.
{"title":"Experimental Validation of Multi-fidelity Models for Prognostics of Electromechanical Actuators","authors":"L. Baldo, P. Berri, M. D. Dalla Vedova, P. Maggiore","doi":"10.36001/phme.2022.v7i1.3347","DOIUrl":"https://doi.org/10.36001/phme.2022.v7i1.3347","url":null,"abstract":"The growing adoption of electrical energy as a secondary form of onboard power leads to an increase of electromechanical actuators (EMAs) use in aerospace applications. Therefore, innovative prognostic and diagnostic methodologies are becoming a fundamental tool to early identify faults propagation, prevent performance degradation, and ensure an acceptable level of safety and reliability of the system. Furthermore, prognostics entails further advantages, including a better ability to plan the maintenance of the various equipment, manage the warehouse and maintenance personnel, and a reduction in system management costs.\u0000Frequently, such approaches require the development of typologies of numerical models capable of simulating the performance of the EMA with different levels of fidelity: monitoring models, suitably simplified to combine speed and accuracy with reduced computational costs, and high fidelity models (and high computational intensity), to generate databases, develop predictive algorithms and train machine learning surrogates. Because of this, the authors developed a high-fidelity multi-domain numerical model (HF) capable of accounting for a variety of physical phenomena and gradual failures in the EMA, as well as a low-fidelity counterpart (LF). This simplified model is derived by the HF and intended for monitoring applications. While maintaining a low computing cost, LF is fault sensitive and can simulate the system position, speed, and equivalent phase currents.\u0000These models have been validated using a dedicated EMA test bench, designed and implemented by authors. The HF model can simulate the operation of the actuator in nominal conditions as well as in the presence of incipient mechanical faults, such as a variation in friction and an increase of backlash in the reduction gearbox.\u0000Comparing the preliminary results highlights satisfactory consistency between the experimental test bench and the two numerical models proposed by the authors.","PeriodicalId":422825,"journal":{"name":"PHM Society European Conference","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121425942","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：