A Bayesian Machine Learning Approach for Efficient Integrity Management of Steel Lazy Wave Risers

2020 ◽  
Author(s):  
Rasoul Hejazi ◽  
Andrew Grime ◽  
Mark Randolph ◽  
Mike Efthymiou
Keyword(s):  
Machine Learning ◽  
Fatigue Failure ◽  
Learning Process ◽  
Predictive Models ◽  
Structural Integrity ◽  
Degradation Mechanism ◽  
Data Driven ◽  
Practical Implementation ◽  
Integrity Management

Abstract In-service integrity management (IM) of steel lazy wave risers (SLWRs) can benefit significantly from quantitative assessment of the overall risk of system failure as it can provide an effective tool for decision making. SLWRs are prone to fatigue failure within their touchdown zone (TDZ). This failure mode needs to be evaluated rigorously in riser IM processes because fatigue is an ongoing degradation mechanism threatening the structural integrity of risers throughout their service life. However, accurately evaluating the probability of fatigue failure for riser systems within a useful time frame is challenging due to the need to run a large number of nonlinear, dynamic numerical time domain simulations. Applying the Bayesian framework for machine learning, through the use of Gaussian Processes (GP) for regression, offers an attractive solution to overcome the burden of prohibitive simulation run times. GPs are stochastic, data-driven predictive models which incorporate the underlying physics of the problem in the learning process, and facilitate rapid probabilistic assessments with limited loss in accuracy. This paper proposes an efficient framework for practical implementation of a GP to create predictive models for the estimation of fatigue responses at SLWR hotspots. Such models are able to perform stochastic response prediction within a few milliseconds, thus enabling rapid prediction of the probability of SLWR fatigue failure. A realistic North West Shelf (NWS) case study is used to demonstrate the framework, comprising a 20” SLWR connected to a representative floating facility located in 950 m water depth. A full hindcast metocean dataset with associated statistical distributions are used for the riser long-term fatigue loading conditions. Numerical simulation and sampling techniques are adopted to generate a simulation-based dataset for training the data-driven model. In addition, a recently developed dimensionality reduction technique is employed to improve efficiency and reduce complexity of the learning process. The results show that the stochastic predictive models developed by the suggested framework can predict the long-term TDZ fatigue damage of SLWRs due to vessel motions with an acceptable level of accuracy for practical purposes.

Risk-Informed Structural Integrity Management: Development of SMR-Centric Piping Reliability Models

2014 ◽  
Author(s):  
Bengt Lydell
Keyword(s):  
Stainless Steel ◽  
Austenitic Stainless Steel ◽  
Structural Integrity ◽  
Degradation Mechanism ◽  
Management Strategies ◽  
Mitigation Strategies ◽  
Parameter Estimates ◽  
Piping System ◽  
Reliability Models ◽  
Integrity Management

In the context of risk-informed applications, this paper addresses the progress with piping reliability analysis methods and techniques and their role in supporting development of risk-informed structural integrity programs for small modular reactors (SMRs). The structural integrity of a pressure boundary is determined by multiple and interrelated reliability attributes and influence factors. Depending on the conjoint requirements for damage and degradation, certain combinations of material, operating environment, loading conditions together with applicable design codes and standards, certain passive components are substantially more resistant to damage and degradation than others. As an example, for stabilized austenitic stainless steel pressure boundary components, there are no recorded events involving active, through-wall leakage. By contrast, for unstabilized austenitic stainless steel, multiple events involving through-wall leakage have been recorded, albeit with relative minor leak rates. The field experience with safety- and non-safety related piping in commercial GenI through GenIII nuclear power reactors is quite extensive. Equally extensive is the experience gained from the implementation of different degradation mechanism mitigation strategies. By applying advanced piping reliability models, this body engineering data and integrity management insights can be used to assess the projected structural integrity of new piping system designs, including those of SMRs. The paper presents an overview of recent methodological advances and insights from the application of statistical piping reliability models to advanced reactor designs. Examples are provided on how piping reliability parameter estimates are affected by different integrity management strategies as well as by advanced, degradation mechanism (DM) resistant materials. The technical basis for the work that is presented in this paper has evolved over a period of 20+ years of focused and sustained R&D in the area of statistical models of piping reliability.

Application of Machine Learning Techniques for Real Time Rate of Penetration Optimization

2021 ◽  
Author(s):  
Asad Mustafa Elmgerbi ◽  
Clemens Peter Ettinger ◽  
Peter Mbah Tekum ◽  
Gerhard Thonhauser ◽  
Andreas Nascimento
Keyword(s):  
Machine Learning ◽  
Genetic Algorithm ◽  
Real Time ◽  
Predictive Models ◽  
Data Driven ◽  
Machine Learning Techniques ◽  
Time Rate ◽  
Rate Of Penetration ◽  
Drilling Parameters ◽  
Learning Techniques

Abstract Over the past decade, several models have been generated to predict Rate of Penetration (ROP) in real-time. In general, these models can be classified into two categories, model-driven (analytical models) and data-driven models (based on machine learning techniques), which is considered as cutting-edge technology in terms of predictive accuracy and minimal human interfering. Nevertheless, most existing machine learning models are mainly used for prediction, not optimization. The ROP ahead of the bit for a certain formation layer can be predicted with such methods, but the limitation of the applications of these techniques is to find an optimum set of operating parameters for the optimization of ROP. In this regard, two data-driven models for ROP prediction have been developed and thereafter have been merged into an optimizer model. The purpose of the optimization process is to seek the ideal combinations of drilling parameters that would lead to an improvement in the ROP in real-time for a given formation. This paper is mainly focused on describing the process of development to create smart data-driven models (built on MATLAB software environment) for real-time rate of penetration prediction and optimization within a sufficient time span and without disturbing the drilling process, as it is typically required by a drill-off test. The used models here can be classified into two groups: two predictive models, Artificial Neural Network (ANN) and Random Forest (RF), in addition to one optimizer, namely genetic algorithm. The process started by developing, optimizing, and validation of the predictive models, which subsequently were linked to the genetic algorithm (GA) for real-time optimization. Automated optimization algorithms were integrated into the process of developing the productive models to improve the model efficiency and to reduce the errors. In order to validate the functionalities of the developed ROP optimization model, two different cases were studied. For the first case, historical drilling data from different wells were used, and the results confirmed that for the three known controllable surface drilling parameters, weight on bit (WOB) has the highest impact on ROP, followed by flow rate (FR) and finally rotation per minute (RPM), which has the least impact. In the second case, a laboratory scaled drilling rig "CDC miniRig" was utilized to validate the developed model, during the validation only the previous named parameters were used. Several meters were drilled through sandstone cubes at different weights on bit, rotations per minute, and flow rates to develop the productive models; then the optimizer was activated to propose the optimal set of the used parameters, which likely maximize the ROP. The proposed parameters were implemented, and the results showed that ROP improved as expected.

Machine Learning in Updating Predictive Models of Planning and Scheduling Transportation Projects

1997 ◽  
Vol 1588 (1) ◽  
pp. 86-94
Author(s):  
Liye Zhang ◽  
W. M. Kim Roddis
Keyword(s):  
Machine Learning ◽  
Data Base ◽  
Learning Process ◽  
Predictive Models ◽  
Functional Unit ◽  
Large Data ◽  
Planning And Scheduling ◽  
Planning Factor ◽  
Functional Units ◽  
Internal Management System

A method combining machine learning and regression analysis to automatically and intelligently update predictive models used in the Kansas Department of Transportation’s (KDOT’s) internal management system is presented. The predictive models used by KDOT consist of planning factors (mathematical functions) and base quantities (constants). The duration of a functional unit (defined as a subactivity) is determined by the product of a planning factor and its base quantity. The availability of a large data base on projects executed over the past decade provided the opportunity to develop an automated process updating predictive models based on extracting information from historical data through machine learning. To perform the entire task of updating the predictive models, the learning process consists of three stages. The first stage derives the numerical relationship between the duration of a functional unit and the project attributes recorded in the data base. The second stage finds the functional units with similar behavior—that is, identifies functional units that can be described by the same shared planning factor scaled in terms of their own base quantities. The third stage generates new planning factors and base quantities. A system called PFactor built on the basis of the three-stage learning process shows good performance in updating KDOT’s predictive models.

scholarly journals Comparison of Empirical Mode Decomposition, Wavelets, and Different Machine Learning Approaches for Patient-Specific Seizure Detection Using Signal-Derived Empirical Dictionary Approach

2021 ◽  
Vol 3 ◽  
Author(s):  
Muhammad Kaleem ◽  
Aziz Guergachi ◽  
Sridhar Krishnan
Keyword(s):  
Machine Learning ◽  
Seizure Detection ◽  
Reconstruction Error ◽  
Data Driven ◽  
Patient Specific ◽  
Learning Approaches ◽  
Mode Decomposition ◽  
Eeg Data ◽  
Classifier Training

Analysis of long-term multichannel EEG signals for automatic seizure detection is an active area of research that has seen application of methods from different domains of signal processing and machine learning. The majority of approaches developed in this context consist of extraction of hand-crafted features that are used to train a classifier for eventual seizure detection. Approaches that are data-driven, do not use hand-crafted features, and use small amounts of patients' historical EEG data for classifier training are few in number. The approach presented in this paper falls in the latter category, and is based on a signal-derived empirical dictionary approach, which utilizes empirical mode decomposition (EMD) and discrete wavelet transform (DWT) based dictionaries learned using a framework inspired by traditional methods of dictionary learning. Three features associated with traditional dictionary learning approaches, namely projection coefficients, coefficient vector and reconstruction error, are extracted from both EMD and DWT based dictionaries for automated seizure detection. This is the first time these features have been applied for automatic seizure detection using an empirical dictionary approach. Small amounts of patients' historical multi-channel EEG data are used for classifier training, and multiple classifiers are used for seizure detection using newer data. In addition, the seizure detection results are validated using 5-fold cross-validation to rule out any bias in the results. The CHB-MIT benchmark database containing long-term EEG recordings of pediatric patients is used for validation of the approach, and seizure detection performance comparable to the state-of-the-art is obtained. Seizure detection is performed using five classifiers, thereby allowing a comparison of the dictionary approaches, features extracted, and classifiers used. The best seizure detection performance is obtained using EMD based dictionary and reconstruction error feature and support vector machine classifier, with accuracy, sensitivity and specificity values of 88.2, 90.3, and 88.1%, respectively. Comparison is also made with other recent studies using the same database. The methodology presented in this paper is shown to be computationally efficient and robust for patient-specific automatic seizure detection. A data-driven methodology utilizing a small amount of patients' historical data is hence demonstrated as a practical solution for automatic seizure detection.

scholarly journals Predictive Models on COVID 19: What Africans Should Do?

2020 ◽  
Author(s):  
Habte Tadesse Likassa
Keyword(s):  
Predictive Models ◽  
Practical Implementation ◽  
The Public ◽  
Daily Report ◽  
Cubic Model ◽  
Cubic Polynomial ◽  
Novel Virus ◽  
The Government ◽  
Global Health Problem

Abstract BackgroundCOVID 19 is becoming a global health problem, where strong intervention is needed. Thus, this paper addresses predictive models on COVID 19 in Africa, from which the government and others put a strong intervention in optimizing resources and necessary healthcare demand.MethodsPredictive models (Cubic polynomial and quadratic regression models) are considered based on the daily report of WHO, 2020 rampant data. The data were analyzed using R and STATA packages.ResultsThe result of the study has shown that the spatial and temporal pattern of this novel virus is varying, spreading and covering the entire world within a brief time. The result has shown that the fitting effect of cubic model is best outperforming compared to the other six families of exponentials ( {R}^{2}=0.996, F=538.334, {D}_{{F}_{1}}=3,{D}_{{F}_{1}}=7, {b}_{1}=13691.949, {b}_{2}=-824.701, {b}_{1}=12.956). The cubic algorithm is more robust in predicting the deaths and confirmed cases of COVID 19. There are also evidences that the source of the outbreak of the epidemic is related to Huanan Seafood from the whole market, fever (78%), cough (59%), fatigue (75%), headache (76%), and others are identified as the major symptoms of COVID 19. Moreover, the result of our study has shown the corona virus infection epidemic is increasing, which seeks a long-term plan to take an action in disease prevention and intervention programs.ConclusionThe trend of COVID 19 is increasing with an alarm rate, thus strong intervention is needed to mitigate the spread of this novel virus. This also can be done through reducing the spread of COVID 19 as persistent and strict self-isolation. The results acquired from this study also recommend that COVID-19 mortality and more cases might be engulfing in Africa due to lack of preparedness and giving strong awareness for the public.This pandemic will sustain to grow up, and peak to the highest for which a strong care and public health interventions practically implemented. Africans must go beyond theory preparations, strong awareness for the public and practical implementation is highly recommended. Highly recommended more sophisticated equipment to tackle the spread of the virus and safe the loss of the infected from deaths.

scholarly journals Machine learning for suicide risk prediction in children and adolescents with electronic health records

2020 ◽  
Vol 10 (1) ◽  
Author(s):  
Chang Su ◽  
Robert Aseltine ◽  
Riddhi Doshi ◽  
Kun Chen ◽  
Steven C. Rogers ◽  
...  
Keyword(s):  
Machine Learning ◽  
Risk Factors ◽  
Children And Adolescents ◽  
Suicidal Behavior ◽  
Predictive Models ◽  
Suicide Risk ◽  
Health Records ◽  
Short And Long Term ◽  
Clinical Records

AbstractAccurate prediction of suicide risk among children and adolescents within an actionable time frame is an important but challenging task. Very few studies have comprehensively considered the clinical risk factors available to produce quantifiable risk scores for estimation of short- and long-term suicide risk for pediatric population. In this paper, we built machine learning models for predicting suicidal behavior among children and adolescents based on their longitudinal clinical records, and determining short- and long-term risk factors. This retrospective study used deidentified structured electronic health records (EHR) from the Connecticut Children’s Medical Center covering the period from 1 October 2011 to 30 September 2016. Clinical records of 41,721 young patients (10–18 years old) were included for analysis. Candidate predictors included demographics, diagnosis, laboratory tests, and medications. Different prediction windows ranging from 0 to 365 days were adopted. For each prediction window, candidate predictors were first screened by univariate statistical tests, and then a predictive model was built via a sequential forward feature selection procedure. We grouped the selected predictors and estimated their contributions to risk prediction at different prediction window lengths. The developed predictive models predicted suicidal behavior across all prediction windows with AUCs varying from 0.81 to 0.86. For all prediction windows, the models detected 53–62% of suicide-positive subjects with 90% specificity. The models performed better with shorter prediction windows and predictor importance varied across prediction windows, illustrating short- and long-term risks. Our findings demonstrated that routinely collected EHRs can be used to create accurate predictive models for suicide risk among children and adolescents.

Auto-control model building using machine learning regression for extreme response prediction

2020 ◽  
Vol 60 (1) ◽  
pp. 155
Author(s):  
Darrell Leong ◽  
Anand Bahuguni
Keyword(s):  
Machine Learning ◽  
Model Building ◽  
Response Prediction ◽  
Offshore Structures ◽  
Practical Implementation ◽  
Monte Carlo Data ◽  
Sea State ◽  
Extreme Load ◽  
Extreme Response

The long-term forecast of extreme response presents a daunting practical problem for offshore structures. These installations are subject to varying sea conditions, which amplify the need to account for the uncertainties of wave heights and periods across a given sea state. Analysis of each sea state involves numerically intensive non-linear dynamic analysis, leading to massive computational expense across the environmental scatter diagram. Recent research has proposed several effective solutions to predict long-term extreme responses, but not without drawbacks, such as the limitation to specific failure locations and the absence of error estimates. This paper explores the practical implementation of control variates as an efficiency enhancing post-processing technique. The model building framework exhibits the advantage of being fully defined from existing simulation results, without the need for external inputs to set up a control experiment. A composite machine learning regression model is developed and investigated for performance in correlating against Monte Carlo data. The sampling methodology presented possesses a crucial advantage of being independent of failure characteristics, allowing for the concurrent extreme response analyses of multiple components across the global structure without the need for re-analysis. The approach is applied on a simulated floating production storage and offloading unit in a site located in the hurricane-prone Gulf of Mexico, vulnerable to heavy-tailed extreme load events.

A Co-Reliability-Target-Based Fatigue Failure Probability Model for Implementing the New ASME Boiler and Pressure Vessel Section XI Division 2 Reliability and Integrity Management Code: A Technical Brief

2019 ◽  
Author(s):  
Jeffrey T. Fong ◽  
N. A. Heckert ◽  
J. J. Filliben ◽  
Stephen W. Freiman
Keyword(s):  
Fatigue Failure ◽  
Pressure Vessel ◽  
Structural Integrity ◽  
Elevated Temperatures ◽  
Probability Model ◽  
Rupture Life ◽  
Creep Rupture ◽  
Time Data ◽  
Integrity Management ◽  
Division 2

Abstract ASME Pressure Vessel and Piping Code (BPVC) Section XI Committee Division 2 has recently completed the development of a proposed new code named the “Reliability and Integrity Management (RIM).” RIM was developed using the system-based code (SBC) concept proposed by the Japanese Society of Mechanical Engineers (JSME) in 2012. Key to the SBC concept is the requirement for the establishment of a co-reliability target (or, reliability-index target by JSME terminology). Such target is usually in the low probability ranges such as 1.0 E−8 to 1.0 E−3. In a paper presented at the 2018 International Symposium on Structural Integrity (ISSI2018), Nov. 2–5, 2018, Nanjing, China, we developed a new theory of fatigue and creep rupture life modeling for metal alloys at room and elevated temperatures such that the co-reliability target can be estimated from fatigue and creep tupture test data. To illustrate an application of this new modeling approach, we included two numerical examples using (a) the fatigue failure data of six AISI 4340 steel specimens at room temperature (Dowling, N. E., 1973) and (b) the creep rupture time data of 37 specimens of 1.3Mn-0.5Mo-0.5Ni steel plates at 500 C (NRIM, 1987). Since the ASME BPVC Section XI Div. 2 RIM Code has just been developed, and the information in that yet-to-be-published ISSI2018 Proceedings is critical to the implementation of that RIM Code, it is, therefore, of interest to the engineering community to have a preview of that information in the form of a technical brief as described in this PVP2019 conference paper.

Machine Learning Application to Estimate Missing Rock Mechanical Logs and Reduce Uncertainty While Drilling

2021 ◽  
Author(s):  
Ardiansyah Negara ◽  
Arturo Magana-Mora ◽  
Khaqan Khan ◽  
Johannes Vossen ◽  
Guodong David Zhan ◽  
...  
Keyword(s):  
Machine Learning ◽  
Real Time ◽  
Predictive Models ◽  
Gamma Ray ◽  
Well Logging ◽  
Complex Model ◽  
Machine Learning Algorithms ◽  
Data Driven ◽  
Time Data ◽  
Drilling Data

Abstract This study presents a data-driven approach using machine learning algorithms to provide predicted analogues in the absence of acoustic logs, especially while drilling. Acoustic logs are commonly used to derive rock mechanical properties; however, these data are not always available. Well logging data (wireline/logging while drilling - LWD), such as gamma ray, density, neutron porosity, and resistivity, are used as input parameters to develop the data-driven rock mechanical models. In addition to the logging data, real-time drilling data (i.e., weight-on-bit, rotation speed, torque, rate of penetration, flowrate, and standpipe pressure) are used to derive the model. In the data preprocessing stage, we labeled drilling and well logging data based on formation tops in the drilling plan and performed data cleansing to remove outliers. A set of field data from different wells across the same formation is used to build and train the predictive models. We computed feature importance to rank the data based on the relevance to predict acoustic logs and applied feature selection techniques to remove redundant features that may unnecessarily require a more complex model. An additional feature, mechanical specific energy, is also generated from drilling real-time data to improve the prediction accuracy. A number of scenarios showing a comparison of different predictive models were studied, and the results demonstrated that adding drilling data and/or feature engineering into the model could improve the accuracy of the models.

Sign in / Sign up

Export Citation Format

Share Document