Data-driven Diagnostics Research Articles

Circuit categorization approach of office building energy consumption based on data features for energy-saving diagnosis

Energy-saving in office buildings is crucial. Research on data-driven diagnostics of building energy consumption is pivotal. The emergence and development of itemized electricity platforms have expanded the scope of diagnosis from total building energy consumption to individual electrical equipment. However current studies rarely address the characteristics of sub-circuits. This paper highlights the importance of recognizing the features of electrical equipment sub-circuits based on historical energy consumption data by examining various aspects of energy use in office buildings, including electrical consumption disaggregation, equipment operational strategies, sub-item consumption proportion, and sub-circuit composition. A circuit categorization approach in office buildings by features analysis of historical data is innovatively proposed. Operational conditions, time series stationarity, and correlations with factors affecting energy consumption are considered in historical data feature analysis. The method categorizes the electric circuits into 5 categories with typical features based on the ADF-KPSS co-test and Spearman correlation analysis. The rationality and validity of the circuit categorization method were verified with actual building data in case studies. On this basis, an interpretable energy consumption anomaly diagnosis method based on the categorization results is proposed. The circuit categorization approach explores how to further investigate the typical features of sub-circuits, and provides new directions for subsequent building operation and maintenance(O&M) research, including energy consumption diagnostics. Additionally, it offers decision support for building managers for O&M evaluation from an interpretable perspective.

Human digital twin technology for individual profiling using LC-MS untargeted metabolomics analysis of dried blood spot samples

Background Digital twins in healthcare enable the creation of accurate, real-time replicas of individual patients, allowing for personalized, data-driven diagnostics, treatment plans, and monitoring to enhance patient outcomes and healthcare efficiency. This transformative shift could significantly enhance our ability to comprehend and address individual well-being and health needs in a more personalized and preventive manner. This study aims to highlight the approach developed by BioTwin Inc., designed to assist and empower healthcare providers in their clinical practice. Methods Through the integration of dried blood samples (DBS), biometric data, untargeted metabolomics liquid chromatography and mass spectrometry (LC-MS) analysis, and a data-driven workflow, BioTwin Inc. holds the potential to generate insights into individual health, potentially catalyzing a transformative shift in healthcare. To achieve this, a cross-sectional study was conducted to collect DBS samples from 277 volunteers over 30 months across Canada and the United States of America. Samples were collected using standardized protocols and analyzed using LC-MS. Subsequently, a machine learning approach was employed for further analysis and refinement of prediction models. Results The results of the experiment demonstrate the dynamic nature of metabolism, revealing its variability within individuals over time and its uniqueness across different individuals. The precision for predicting sample ownership was 80% accuracy when users provided 5 samples and 92% accuracy when users provided 10 samples. These findings underscore the importance of understanding temporal variations and individuality in metabolomics research. Conclusions In conclusion, the use of digital twins in healthcare, coupled with untargeted metabolomics and advanced machine learning techniques, has the potential to revolutionize healthcare delivery. The emphasis on individuality, dynamic metabolic profiles, and precision in patient care opens new frontiers in personalized and preventive medicine. Moving forward, the integration of multiple data sources and the synergy between metabolic and biometric data will benefit both providers and patients.

Artificial intelligence and edge computing for machine maintenance-review

Industrial internet of things (IIoT) has ushered us into a world where most machine parts are now embedded with sensors that collect data. This huge data reservoir has enhanced data-driven diagnostics and prognoses of machine health. With technologies like cloud or centralized computing, the data could be sent to powerful remote data centers for machine health analysis using artificial intelligence (AI) tools. However, centralized computing has its own challenges, such as privacy issues, long latency, and low availability. To overcome these problems, edge computing technology was embraced. Thus, instead of moving all the data to the remote server, the data can now transition on the edge layer where certain computations are done. Thus, access to the central server is infrequent. Although placing AI on edge devices aids in fast inference, it poses new research problems, as highlighted in this paper. Moreover, the paper discusses studies that use edge computing to develop artificial intelligence-based diagnostic and prognostic techniques for industrial machines. It highlights the locations of data preprocessing, model training, and deployment. After analysis of several works, trends of the field are outlined, and finally, future research directions are elaborated

Data-driven diagnostics of boiling heat transfer on flat heaters from non-intrusive visualization

Rankine-based power generation cycles require the use of boiling heat exchangers for liquid–vapor phase change. However, operating boilers at too high heat fluxes can lead to heater meltdown due to onset of the boiling crisis. Despite recent advances in increasing this critical heat flux, the lack of fast and accurate diagnostics hinders boiling heat transfer safety at high heat fluxes. Here we show that neural networks fed on direct observation of bubble dynamics can be used to measure heat flux during nucleate boiling. The boiling heat transfer data was obtained with a horizontally oriented, circular flat heater with saturated water as the heat transfer fluid. A deep convolutional neural network (CNN) was trained using frames from videos acquired from 15 different heat dissipation values on the boiling curve to learn how these images correlate to the underlying heat flux. The CNN predicted heat fluxes with a mean square error of 14.6 W/cm2. Furthermore, the CNN presented high sensitivity to pixel intensity in image regions relevant to boiling heat transfer physics (e.g. bubble boundaries, contextual clues about bubble image distortion). Understanding this approach may allow future studies in unstable and critical boiling regimes that require fast transient heat flux diagnostics and cannot rely on slower thermohydraulic measurements.

Statistical Models for Condition Monitoring and State of Health Estimation of Lithium-Ion Batteries for Ships

Battery systems are increasingly being used for powering ocean going ships, and the number of fully electric or hybrid ships relying on battery power for propulsion is growing. To ensure the safety of such ships, it is important to monitor the available energy that can be stored in the batteries, and classification societies typically require the state of health (SOH) to be verified by independent tests. This paper addresses statistical modelling of SOH for maritime lithium-ion batteries based on operational sensor data. Various methods for sensor-based, data-driven degradation monitoring will be presented, and advantages and challenges with the different approaches will be discussed. The different approaches include cumulative degradation models and snapshot models, models that need to be trained and models that need no prior training, and pure data-driven models and physics-informed models. Some of the methods only rely on measured data, such as current, voltage and temperature, whereas others rely on derived quantities such as state of charge (SOC). Models include simple statistical models and more complicated machine learning techniques. Insight from this exploration will be important in establishing a framework for data-driven diagnostics and prognostics of maritime battery systems within the scope of classification societies. Conflict of Interest Statement The authors declare no conflicts of interest.

Data-Driven Approaches to Diagnostics and State of Health Monitoring of Maritime Battery Systems

Battery systems are increasingly being used for powering ocean going ships, and the number of fully electric or hybrid ships relying on battery power for propulsion and maneuvering is growing. In order to ensure the safety of such electric ships, it is important to monitor the available energy that can be stored in the batteries, and classification societies typically require that the state of health (SOH) can be verified by independent tests. However, this paper addresses data-driven approaches to state of health monitoring of maritime battery systems based on operational sensor data. Results from various approaches to sensor-based, data-driven degradation monitoring of maritime battery systems will be presented, and advantages and challenges with the different methods will be discussed. The different approaches include cumulative degradation models and snapshot models. Some of the models need to be trained, whereas others need no prior training. Moreover, some of the methods only rely on measured data, such as current, voltage and temperature, whereas others rely on derived quantities such as state of charge (SOC). Models include simple statistical models and more complicated machine learning techniques. Different datasets have been used in order to explore the various methods, including public datasets, data from laboratory tests and operational data from ships in actual operation. Lessons learned from this exploration will be important in establishing a framework for data-driven diagnostics and prognostics of maritime battery systems within the scope of classification societies.

Modeling the Tokamak Exhaust Processing System in a Commercial Simulator for Process Monitoring Purposes

Nuclear fusion depends on tritium breeding and self-sufficiency. Tritium represents a hazard due to its radioactivity and migration properties. Because of these difficulties, ITER, the largest fusion experiment so far, relies on a conservative static procedure to monitor the tritium inventory. Future commercial fusion plants can avoid operation halts if a dynamic monitoring strategy proves itself valid. Tritium plant models have been developed for this kind of monitoring and analysis task, but sensor accuracy and reliability are an issue still to be addressed, and the path to dynamic monitoring remains unclear. The present work shows the modeling procedure of the Tokamak Exhaust Processing system in a commercial simulator, Aspen HYSYS, to reproduce the inventories, streams, process conditions, and compositions of this subsystem during operation. The model is verified in a steady-state scenario using data from the available literature. A demonstration of such a tritium plant subsystem shows meaningful value for several reasons. First, this process has not been modeled before in commercial dynamic simulators, which are typically used in the process industry. It will also allow new stakeholders to participate in future fusion-related projects. Second, it will play a key role in industry-like tritium process monitoring, in which the new model will act as a digital twin of the plant. Data-driven diagnostics can be fueled by model data, helping engineers to generate additional data that could otherwise be expensive to get directly from the plant. For these reasons, models will represent an essential part of a dynamic monitoring system, necessary for feasible fusion projects.

Hybrid deep fault detection and isolation: Combining deep neural networks and system performance models

With the increased availability of condition monitoring data on the one hand and the increased complexity of explicit system physics-based models on the other hand, the application of data-driven approaches for fault detection and isolation has recently grown. While detection accuracy of such approaches is generally very good, their performance on fault isolation often suffers from the fact that fault conditions affect a large portion of the measured signals thereby masking the fault source. To overcome this limitation, we propose a hybrid approach combining physical performance models with deep learning algorithms. Unobserved process variables are inferred with a physics-based performance model to enhance the input space of a data-driven diagnostics model. The resulting increased input space gains representation power enabling more accurate fault detection and isolation. To validate the effectiveness of the proposed method, we generate a condition monitoring dataset of an advanced gas turbine during flight conditions under healthy and four faulty operative conditions based on the Aero-Propulsion System Simulation (C-MAPSS) dynamical model. We evaluate the performance of the proposed hybrid methodology in combination with two different deep learning algorithms: deep feed forward neural networks and Variational Autoencoders, both of which demonstrate a significant improvement when applied within the hybrid fault detection and diagnostics framework. The proposed method is able to outperform pure data-driven solutions, particularly for systems with a high variability of operating conditions. It provides superior results both for fault detection as well as for fault isolation. For the fault isolation task, it overcomes the smearing effect that is commonly observed in pure data-driven approaches and enables a precise isolation of the affected signal. We also demonstrate that deep learning algorithms provide a better performance on the fault detection task compared to the traditional machine learning algorithms.

Data-Driven Diagnostics: Leveraging Machine Learning for Precision Medicine

The field of precision medicine holds great promise in revolutionizing healthcare by providing personalized and targeted treatments. However, the successful implementation of precision medicine heavily relies on accurate and efficient diagnostic tools. This paper explores the potential of leveraging machine learning techniques for data-driven diagnostics in precision medicine

Informed machine learning for image-data-driven diagnostics of hydrogenerators

Visual inspection is one of the most common and reliable methods used by human experts to perform diagnostics in the industry. However, it requires costly, specific expertise that could benefit from being automated. Such specific tasks are precisely the type of narrow knowledge that machine learning algorithms are best at learning. Yet, data-driven diagnostics from images are lagging behind compared to that from sensor data. This is due to sparse image data, as human experts only need a few photographs to accurately diagnose machine degradation. This paper presents a methodology to incorporate expert knowledge into the development of a data-driven diagnostic model for hydrogenerators based on visual inspection of the presence of partial discharge degradation products. The proposed methodology is validated using real industrial images. It emphasizes on the integration of human knowledge to address multiple challenges such as data sparsity, knowledge conformity, and human interpretability.

Data-Driven Diagnostics Based on Non-invasive Monitoring Using Electrical Signals: Application to Rotating Machines

Data-Driven Diagnostics Based on Non-invasive Monitoring Using Electrical Signals: Application to Rotating Machines

A Data-Driven Snapshot Method for State of Health Modelling and Diagnostics of Maritime Battery Systems

Battery systems are increasingly being used for powering ocean going ships, and the number of fully electric or hybrid ships relying on battery power for propulsion and manoeuvring is growing. In order to ensure the safety of such electric ships, it is of paramount importance to monitor the available energy that can be stored in the batteries, and classification societies typically require that the state of health of the batteries can be verified by independent tests - annual capacity tests. However, this paper discusses data-driven diagnostics for state of health modelling for maritime battery systems based on operational sensor data collected from the batteries as an alternative approach. There are different strategies for such data-driven diagnostics. Some approaches, referred to as cumulative damage models, require full operational history of the batteries in order to predict state of health, and this may be impractical due to several reason. Thus, snapshot methods that are able to give reliable estimation of state of health based on only snapshots of the data streams are attractive candidates for data-driven diagnostics of battery systems on board ships. In this paper, data-driven snapshot methods are explored and applied to a novel set of degradation data from battery cells cycled in laboratory tests. The paper presents the laboratory tests, the resulting battery data, shows how relevant features can be extracted from snapshots of the data and presents data-driven models for state of health prediction. It is discussed how such methods could be utilized in a data-driven classification regime for maritime battery systems. Results are encouraging, and yields reasonable degradation estimates for 40% of the tested cells. This is greatly improved if data from the actual cell is included in the training data, and indicates that better results can be achieved if more representative training data is available. Nevertheless, improved accuracy is required for such snapshot methods to be recommended for ships in actual operation.

A Preliminary Experimental Study on the Effects of Wear on the Torsional Stiffness of Strain Wave Gears

Strain wave gears, also known as harmonic drives, are employed in a wide range of fields such as robotics and aerospace, where light weights, precision, and reliability are essential to the correct execution of the tasks. For this reason, their understanding and optimization are of high interest for both academia and industry. Previous studies have been mainly focused on investigating and modeling the working principle of strain wave gears in nominal operating conditions. On the contrary, the present paper describes the results of an experimental campaign aimed to introduce wear in gears of two different suppliers and its impact on the gear torsional stiffness. Results show how the change in the gear performance strongly depends both on the gear manufacturer and the location of wear. For the analyzed components, a damaged wave generator–flexspline interface reduces the gear stiffness up to one-fourth of its nominal value, while the non-nominal shape of the teeth jeopardizes the gearbox performance, leading up to just 4% of the nominal stiffness values, and resulting in backlash. Such data can be used to properly model the presence of wear in strain wave gears and to train data-driven diagnostics and prognostics routines to effectively detect such a fault.

On the Integration of Fundamental Knowledge about Degradation Processes into Data-Driven Diagnostics and Prognostics Using Theory-Guided Data Science

In Prognostics and Health Management, there are three main approaches for implementing diagnostic and prognostic applications. These approaches are data-driven methods, physical model-based methods, and combinations of them, in the form of hybrid methods. Each of them has specific advantages but also limitations for their purposeful implementation. In the case of data-driven methods, one of the main limitations is the availability of sufficient training data that adequately cover the relevant state space. For model-based methods, on the other hand, it is often the case that the degradation process of the considered technical system is of significant complexity. In such a scenario physics-based modeling requires great effort or is not possible at all. Combinations of data-driven and model-based approaches in form of hybrid approaches offer the possibility to partially mitigate the shortcomings of the other two approaches, however, require a sufficiently detailed data-driven and physics-based model. This paper addresses the transitional field between data-driven and hybrid approaches. Despite the issues of formulating a physics-based model that provides a representation of the degradation process, basic knowledge of the considered system and of the laws governing its degradation process is usually available. Integration of such knowledge into a machine learning process is part of a research field that is either called theory-guided data science, (physics) informed machine learning, physics-based learning or physics guided machine learning. First, the state of research in Prognostics and Health Management on methods of this field is presented and existing research gaps are outlined. Then, a concept is introduced for incorporating fundamental knowledge, such as monotonicity constraints, into data-driven diagnostic and prognostic applications using approaches from theory-guided data science. A special aspect of this concept is its cross-application usability through the consideration of knowledge that repeatedly occurs in diagnostics and prognostics. This is, for example, knowledge about physically justified boundaries whose compliance makes a prediction of the data-driven model plausible in the first place.

Data Driven Seal Wear Classifications using Acoustic Emissions and Artificial Neural Networks

The work presented in this paper is built on a series of experiments aiming to develop a data-driven and automated method for seal diagnostics using Acoustic Emission (AE) features. Seals in machineries operate in harsh conditions, and seal wear in hydraulic cylinders results in fluid leakage, and instability of the piston rod movement. Therefore, regular inspection of seals is required using automated approaches to improve productivity and to reduce unscheduled maintenance. In this study, we implemented a data-driven diagnostics approach which utilizes AE measurements along with light weight Artificial Neural Networks (ANN) as a classifier to investigate the performance and resources (hardware & software) required for implementing a real-time soft sensor unit for monitoring seal wear condition. We used a feedforward multilayer perceptron ANN (Scaled Conjugate Gradient- SCG algorithm) that is trained with the back propagation algorithm, which is a popular network architecture for a multitude of applications (automotive, oil and gas, electronics). We benchmark the developed method against previous work conducted based on Support Vector Machine (SVM), and we compare ANN performance in classifying the running condition of seals in hydraulic cylinders by applying it to both raw (full frequency spectrum) and down sampled frequency measurements. The experiments were performed at varying pressure conditions on a hydraulic test rig that can simulate fluid leakage conditions like that of hydraulic cylinders. The test cases were generated with seals of three different conditions (unworn, semi-worn, worn). From the AE spectrum, the frequency bands were identified with peak power and by heterodyning the signal. This technique results in 10X down sampling without losing the information of interest. Further, the signal was divided into smaller “snapshots” to facilitate rapid diagnosis. In these tests, the diagnosis was made on short-time windows, as low as 0.3 seconds in length. A general set of 16 time and frequency domain features were designed. Then a training set was developed using relevant set of features (4, 5, and 16 features). The data was used to train the ANN (70% training – 30% test & validation) and SVM (60 % training - 40% test and validation). Classification of down sampled measurements, both ANN and SVM were able to accurately classify the status irrespective of the pressure conditions, with an accuracy of ~99% with execution time less than seconds. Therefore, the proposed approach can be applied as part of an automated seal wear classification technique based on AE and ANN/SVM and can be used for real-time monitoring of seal wear in hydraulic cylinders.

Data-Driven Diagnostics and Prognostics for Modelling the State of Health of Maritime Battery Systems – a Review

Battery systems are becoming an increasingly attractive alternative for powering ocean going ships, and the number of fully electric or hybrid ships relying on battery power for propulsion and maneuvering is growing. In order to ensure the safety of such electric ships, it is of paramount importance to monitor the available energy that can be stored in the batteries, and classification societies typically require that the state of health of the batteries can be verified by independent tests – annual capacity tests. However, this paper discusses data-driven diagnostics for state of health modelling for maritime battery systems based on operational sensor data collected from the batteries as an alternative approach. Thus, this paper presents a comprehensive review of different data-driven approaches to state of health modelling, and aims at giving an overview of current state of the art. Furthermore, the various methods for data-driven diagnostics are categorized in a few overall approaches with quite different properties and requirements with respect to data for training and from the operational phase. More than 300 papers have been reviewed, most of which are referred to in this paper. Moreover, some reflections and discussions on what types of approaches can be suitable for modelling and independent verification of state of health for maritime battery systems are presented.

Coupling Machine Learning and High Throughput Multiplex Digital PCR Enables Accurate Detection of Carbapenem-Resistant Genes in Clinical Isolates.

Rapid and accurate identification of patients colonised with carbapenemase-producing organisms (CPOs) is essential to adopt prompt prevention measures to reduce the risk of transmission. Recent studies have demonstrated the ability to combine machine learning (ML) algorithms with real-time digital PCR (dPCR) instruments to increase classification accuracy of multiplex PCR assays when using synthetic DNA templates. We sought to determine if this novel methodology could be applied to improve identification of the five major carbapenem-resistant genes in clinical CPO-isolates, which would represent a leap forward in the use of PCR-based data-driven diagnostics for clinical applications. We collected 253 clinical isolates (including 221 CPO-positive samples) and developed a novel 5-plex PCR assay for detection of blaIMP, blaKPC, blaNDM, blaOXA-48, and blaVIM. Combining the recently reported ML method “Amplification and Melting Curve Analysis” (AMCA) with the abovementioned multiplex assay, we assessed the performance of the AMCA methodology in detecting these genes. The improved classification accuracy of AMCA relies on the usage of real-time data from a single-fluorescent channel and benefits from the kinetic/thermodynamic information encoded in the thousands of amplification events produced by high throughput real-time dPCR. The 5-plex showed a lower limit of detection of 10 DNA copies per reaction for each primer set and no cross-reactivity with other carbapenemase genes. The AMCA classifier demonstrated excellent predictive performance with 99.6% (CI 97.8–99.9%) accuracy (only one misclassified sample out of the 253, with a total of 160,041 positive amplification events), which represents a 7.9% increase (p-value <0.05) compared to conventional melting curve analysis. This work demonstrates the use of the AMCA method to increase the throughput and performance of state-of-the-art molecular diagnostic platforms, without hardware modifications and additional costs, thus potentially providing substantial clinical utility on screening patients for CPO carriage.

A Cost-Aware DNN-Based FDI Technology for Solenoid Pumps

Fluid Pumps serve a critical function in hydraulic and thermodynamic systems, and this often exposes them to prolonged use, leading to fatigue, stress, contamination, filter clogging, etc. On one hand, vibration monitoring for hydraulic components has shown reliable efficiencies in fault detection and isolation (FDI) practices. On the other hand, signal processing techniques provide reliable FDI parameters for artificial intelligence (AI)-based data-driven diagnostics (and prognostics) and have recently attracted global interest across different disciplines and applications. Particularly for cost-aware systems, the choice of diagnostic parameters determines the reliability of an FDI/diagnostic model. By extracting (and selecting) discriminative spectral and transient features from solenoid pump vibration signals, accurate diagnostics across operating conditions can be achieved using AI-based FDI algorithms. This study employs a deep neural network (DNN) for fault diagnosis after a correlation-based selection of discriminative spectral and transient features. To solve the problem of hyperparameter selection for the proposed model, a grid search technique was employed for optimal search for parameters (number of layers, neurons, activation function, weight optimizer, etc.) on different network architectures.The results reveal the high accuracy of a three-layer DNN with ReLU activation function, with a test accuracy of 99.23% and a minimal false alarm rate on a case study.

Health Monitoring of Aerospace Structures Utilizing Novel Health Indicators Extracted from Complex Strain and Acoustic Emission Data.

The development of health indicators (HI) of diagnostic and prognostic potential from generally uninformative raw sensor data is both a challenge and an essential feature for data-driven diagnostics and prognostics of composite structures. In this study, new damage-sensitive features, developed from strains acquired with Fiber Bragg Grating (FBG) and acoustic emission (AE) data, were investigated for their suitability as HIs. Two original fatigue test campaigns (constant and variable amplitude) were conducted on single-stringer composite panels using appropriate sensors. After an initial damage introduction in the form of either impact damage or artificial disbond, the panels were subjected to constant and variable amplitude compression–compression fatigue tests. Strain sensing using FBGs and AE was employed to monitor the damage growth, which was further verified by phased array ultrasound. Several FBGs were incorporated in special SMARTapesTM, which were bonded along the stiffener’s feet to measure the strain field, whereas the AE sensors were strategically placed on the panels’ skin to record the acoustic emission activity. HIs were developed from FBG and AE raw data with promising behaviors for health monitoring of composite structures during service. A correlation with actual damage was attempted by leveraging the measurements from a phased array camera at several time instances throughout the experiments. The developed HIs displayed highly monotonic behaviors while damage accumulated on the composite panel, with moderate prognosability.

Application of improved data-driven diagnostics workflow based on coupled hydro-mechanical model for stimulation candidate selection in a complex carbonate reservoir

The focus of the present paper is on utilizing a comprehensive diagnostics workflow that combines coupled hydro-mechanical modeling with production data-driven diagnostics for optimization of stimulation candidate selection process. Reservoir fluids production and production-induced depletion affect reservoir mechanical environment due to reduced pore pressure and increased effective stress and the consequent porosity and permeability reduction, depending on rock sensitivity to stress changes. The coupled formation damage index (CFDI) is implemented in the traditional stimulation candidate selection workflow to capture the effect of production-induced stress changes on the near-wellbore permeability over time. The top potential stimulation candidates are recognized based on heterogeneity index, static formation damage index, and the CFDI parameter. CFDI is a dynamic parameter for stimulation candidate selection through estimating time-dependent permeability changes induced by stress state and fluid pressure. Probabilistic type curves and decline curve analysis of candidate wells versus reservoir unit are also applied to complement the stimulation and production enhancement candidate selection.