Condition Monitoring Applications Research Articles

Bridging the gap between AI and the industry — A study on bearing fault detection in PMSM-driven systems using CNN and inverter measurement

Deep-learning-based identification and diagnosis of faults in industrial assets has received much attention. It does not need a deep understanding of the target domain or complex signal-processing techniques for time-consuming feature identification and selection. The vast majority of approaches developed in this field rely on public data recorded in laboratory conditions with high-performance measurement equipment. Moreover, primarily vibration data is used for condition monitoring tasks, which is particularly sensitive to typical error patterns in rotating machinery. These conditions are difficult to maintain in industrial environments since using high-performance measurement systems would not be economically feasible. This paper demonstrates that modern deep learning can achieve extraordinary fault detection results with industrial standard low-cost measurement systems. We determine measurements taken from the industrial-sensory equipment that potentially capture the system’s fault patterns. Hereafter, we apply traditional failure detection and identification techniques to evaluate the suitability of the taken measurements for bearing fault detection. These techniques apply envelope analysis to reveal the fault patterns in the signals and a support vector machine to separate the fault classes. We can conclude that the used low-cost sensory equipment is capable to capture meaningful, fault-describing patterns. To improve the failure detection performance we are investigating a 2D, as well as 1D, convolutional neural network approach to identify the error patterns and classify the respective errors. We compare the deep-learning-based methods with the traditional methods. Furthermore, we assess which inverter signal carries the largest fault-describing information content. Experimental results indicate that the proposed deep learning methods outperform traditional fault diagnosis methods, hence, demonstrating the effectiveness in an industrial condition monitoring application.

A Mechanical-Electromagnetic Coupling Model of Transformer Windings and its Application in the Vibration-Based Condition Monitoring

Winding vibration leads to the oscillation of coils position and further affects winding's nonlinear stiffness as well as electromagnetic forces (EMFs). This paper establishes a winding Mechanical-Electromagnetic Coupling Model (MECM) where the coupling between stiffness coefficient and vibration, magnetic field and vibration are considered, and deploy it in the vibration-based condition monitoring. In this model, EMFs are obtained through the superposition of leakage magnetic field generated by each coil. A novel numerical analysis method is further proposed, which is capable of calculating the vibration response containing nonlinearity and electromechanical coupling, and the accuracy of the method is verified in experiments. Then, the numerical analysis method is used to quantitatively analyze the effects of Nonlinearity and ElectroMechanical Coupling (NEMC) on the vibration characteristics. The study demonstrates the law that NEMC intensity increases with vibration, and the NEMC is obvious especially under mechanical fault. To extract the vibration patterns, Nonlinearity-ElectroMechanical Coupling Coefficient (NEMCC) with amplifying NEMC effects is proposed to monitor the winding mechanical condition. The analytical method predicts the variation trend of the winding looseness condition curve, and laboratory experiments prove that the Mean Absolute Percentage Error (MAPE) is able to successfully identify different degrees of winding looseness faults. Compared with the short-circuit impedance (SCI) method, the MAPE value based on NEMCC has a stronger detection capability and higher sensitivity for winding mechanical faults. This paper lays a foundation for the numerical solution of vibration response and pave a way for winding mechanical condition monitoring.

Motor Current-Based Degradation Modeling for Tool Wear Hybrid Prognostics in Turning Process

For many machines with turning process systems, the application of economical indirect Tool Condition Monitoring (TCM) is enhanced by utilizing internal encoder spindle motor current signals. In this study, we proposed a novel approach to extract the total harmonic distortion (THD) feature associated with the metal cutting frequency of a specific working tool in the time domain. Our method entailed the application of filtered variational mode decomposition (VMD) combined with envelope analysis to demodulate the motor current signal and define TCM features based on the THD of odd harmonics, which are more related to the motor structure. These features serve as inputs for a hybrid prognostics technique, employing the Geometric Brownian Motion (GBM) to stochastically model the degradation process along with a deep learning transformer-based framework called the time series Transformer (TST) to improve the life prediction. Finally, to validate our approach, we conducted experiments based on 36 sets of tool run-to-wear data extracted from a CNC machine operating under turning process conditions using two different tools. Finally, we compared the degradation models based on the extracted odd-THD and even-THD features.

Explaining deep neural networks processing raw diagnostic signals

Neural networks (NN) have spurred significant interest in automating machine condition monitoring, with many recent studies focusing on networks that take raw diagnostic signals as input. However, the complexity of these NNs has raised concerns about their physical justification and trustworthiness. Although explainable artificial intelligence (XAI) has developed promising tools to address these issues, the nature and size of raw diagnostic inputs have limited the effectiveness of XAI techniques in condition monitoring.This paper proposes a novel approach combining domain transformation and discretisation techniques with the XAI technique of Shapely value additive explanation (SHAP). This methodology allows the input domain for the network to remain in the original time-domain, while the explanation domain can be chosen as frequency or time–frequency and its discretisation adapted to the specific diagnostic case. This flexibility makes networks trained on raw signals more interpretable, while the controllable domain discretisation facilitates the implementation of XAI with manageable computational costs. The proposed methodology is tested on numerical cases and experimental signals, including the UNSW gear-wear and CWRU bearing-fault datasets. This enhancement of XAI is applicable to common black-box classifiers, providing researchers with an effective tool to give physical justification to their models and bolster trust in AI for machine condition monitoring applications.

Assessment of background noise properties in time and time–frequency domains in the context of vibration-based local damage detection in real environment

Any measurement in condition monitoring applications is associated with disturbing noise. Till now, most of the diagnostic procedures have assumed the Gaussian distribution for the noise. This paper shares a novel perspective to the problem of local damage detection. The acquired vector of observations is considered as an additive mixture of signal of interest (SOI) and noise with strongly non-Gaussian, heavy-tailed properties, that masks the SOI. The distribution properties of the background noise influence the selection of tools used for the signal analysis, particularly for local damage detection. Thus, it is extremely important to recognize and identify possible non-Gaussian behavior of the noise. The problem considered here is more general than the classical goodness-of-fit testing. The paper highlights the important role of variance, as most of the methods for signal analysis are based on the assumption of the finite-variance distribution of the underlying signal. The finite variance assumption is crucial but implicit to most indicators used in condition monitoring (such as the root-mean-square value, the power spectral density, the kurtosis, the spectral correlation, etc.), in view that infinite variance implies moments higher than 2 are also infinite. The problem is demonstrated based on three popular types of non-Gaussian distributions observed for real vibration signals. We demonstrate how the properties of noise distribution in the time domain may change by its transformations to the time–frequency domain (spectrogram). Additionally, we propose a procedure to check the presence of the infinite-variance of the background noise. Our investigations are illustrated using simulation studies and real vibration signals from various machines.

Intelligent safe operation and maintenance of oil and gas production systems: Connotations and key technologies

Oil and gas production systems have the characteristics of high operation and maintenance risk and great accident influence. With the deep integration of informationization and industrialization, the development direction and necessary choice of the oil and gas industry is to develop the oil and gas production system into the interconnected, multi-domain interactive cyber-physical intelligent system. In order to avoid or reduce the complex, diverse and potentially unknown safety risks in the process of oil and gas production, improve the safety and reliability of oil and gas production system and increase the production efficiency, this paper analyzes the safety problems occurring in the intelligentization process of oil and gas production system and constructs a system from the perspective of operation and maintenance based on key elements of intelligent safe operation and maintenance technology, combined with the typical production scenarios in the oil and gas production industry. And the following research results are obtained. First, the connotation of intelligent safe operation and maintenance technology is clarified, the key elements and existing problems and challenges of intelligent safe operation and maintenance technology are analyzed, and the “1-2-3-4-5-6” intelligent safe operation and maintenance technology system of oil and gas production system is constructed, which empowers six key technologies with key elements of oil and gas production to realize the essential safety of oil and gas production system. Second, the intelligent safe operation and maintenance technology actively promotes the application and implementation of condition monitoring, health management, risk assessment, intelligent early warning technologies in typical production scenarios such as drilling and extraction, storage and transportation, refining and chemical industry in up, middle and down streams of oil and gas production. Third, in view of the characteristics of oil and gas production system under digital transformation, it is proposed to develop the intelligent safe operation and maintenance technology with the functions of intelligent decision-making, active prevention and comprehensive safety in the future to help the safe construction in the field of oil and gas production and promote the safe and healthy development of the oil and gas industry. In conclusion, the research on intelligent safe operation and maintenance technology system of oil and gas production system is conducive to the safe construction in the field of oil and gas production, which will not only provide technical support for the realization of trouble-free oil and gas production system, but also provide reference for the intelligent development of the world oil and gas industry.

On the Stability and Homogeneous Ensemble of Feature Selection for Predictive Maintenance: A Classification Application for Tool Condition Monitoring in Milling.

Feature selection (FS) represents an essential step for many machine learning-based predictive maintenance (PdM) applications, including various industrial processes, components, and monitoring tasks. The selected features not only serve as inputs to the learning models but also can influence further decisions and analysis, e.g., sensor selection and understandability of the PdM system. Hence, before deploying the PdM system, it is crucial to examine the reproducibility and robustness of the selected features under variations in the input data. This is particularly critical for real-world datasets with a low sample-to-dimension ratio (SDR). However, to the best of our knowledge, stability of the FS methods under data variations has not been considered yet in the field of PdM. This paper addresses this issue with an application to tool condition monitoring in milling, where classifiers based on support vector machines and random forest were employed. We used a five-fold cross-validation to evaluate three popular filter-based FS methods, namely Fisher score, minimum redundancy maximum relevance (mRMR), and ReliefF, in terms of both stability and macro-F1. Further, for each method, we investigated the impact of the homogeneous FS ensemble on both performance indicators. To gain broad insights, we used four (2:2) milling datasets obtained from our experiments and NASA's repository, which differ in the operating conditions, sensors, SDR, number of classes, etc. For each dataset, the study was conducted for two individual sensors and their fusion. Among the conclusions: (1) Different FS methods can yield comparable macro-F1 yet considerably different FS stability values. (2) Fisher score (single and/or ensemble) is superior in most of the cases. (3) mRMR's stability is overall the lowest, the most variable over different settings (e.g., sensor(s), subset cardinality), and the one that benefits the most from the ensemble.

A Time-Series Driven Mechanical System State Description Method and Its Application in Condition Monitoring

Nonlinear and time-varying characteristics of mechanical systems may lead to the variation of fault modes which makes it hard to be accurately evaluated. To fulfill this requirement, the authors propose a time-domain signal-driven mechanical system state description method in this research. Based on the theory of stochastic subspace identification, an abnormal evaluation model is established to investigate the performance of mechanical systems. Firstly, by constructing the projection matrix of the established Hankel matrixes from the given time series at different times, the system state matrix can be obtained. Subsequently, based on the eigenvectors of the reference state matrix, the testing samples can be decomposed. Finally, by calculating the difference between the reference and testing signal, the anomaly can be acquired, and the quantitative evaluation of the health of the mechanical system can also be realized. The proposed method is validated in bearing performance degradation, bearing fault diagnosis, and milling tool wear experiments and received positive results. The proposed method, therefore, provides a new possibility for monitoring the service performance of mechanical systems.

Machine learning explanations by design: a case study explaining the predicted degradation of a roto-dynamic pump

The field of explainable Artificial Intelligence (AI) has gained growing attention over the last few years due to the potential for making accurate data-based predictions of asset health. One of the current research aims in AI is to address challenges associated with adopting machine learning (ML) (i.e., data-driven) AI ‐ that is, understanding how and why ML predictions are made. Despite ML models successfully providing accurate predictions in many applications, such as condition monitoring, there are still concerns about the transparency of the prediction-making process. Therefore, ensuring that the models used are explainable to human users is essential to build trust in the approaches proposed. Consequently, AI and ML practitioners need to be able to evaluate any available eXplainable AI (XAI) tools’ suitability for their intended domain and end users, while simultaneously being aware of the tools’ limitations. This paper provides insight into various existing XAI approaches and their limitations to be considered by practitioners in condition monitoring applications during the design process for an MLbased prediction. The aim is to assist practitioners in engineering applications in building interpretable and explainable models intended for end users who wish to improve a system’s reliability and help users make better-informed decisions based upon a predictive ML algorithm output. It also emphasizes the importance of explainability in AI. The paper applies some of these tools to an explainability use case in which real condition monitoring data is used to predict the degradation of a roto-dynamic pump. Additionally, potential avenues are explored to enhance the credibility of explanations generated by XAI tools in condition monitoring applications, aiming to offer more reliable explanations to domain experts.

On workflow scheduling for latency-sensitive edge computing applications

With the rapid proliferation of edge computing-based solutions, many edge computing applications use the cloud for data processing and analysis. However, latency-sensitive applications have low latency requirements and can be very bandwidth hungry, so processing their collected data through cloud servers is not efficient and cost-effective. For example, traffic management and health condition monitoring applications, require real-time data processing at the network edge to respond immediately to unexpected events. These applications are typically executed as workflows with dependent tasks that require careful scheduling to allocate the appropriate resources to the tasks so that the execution is done in a way that satisfies the users’ target functions. In this work, we evaluate some traditional scheduling heuristics for the execution of workflow tasks in an edge-computing scenario. Our goal is to compare their performance in terms of execution time and cost and show that these heuristics, previously used in scheduling in the context of cloud environments, can also be used in edge computing scenarios. The results show that the MinMin and PSO scheduling algorithms offer the best results with regard to execution time and cost.

Deep Learning Based Underground Sewer Defect Classification Using a Modified RegNet

The sewer system plays an important role in protecting rainfall and treating urban wastewater. Due to the harsh internal environment and complex structure of the sewer, it is difficult to monitor the sewer system. Researchers are developing different methods, such as the Internet of Things and Artificial Intelligence, to monitor and detect the faults in the sewer system. Deep learning is a promising artificial intelligence technology that can effectively identify and classify different sewer system defects. However, the existing deep learning based solution does not provide high accuracy prediction and the defect class considered for classification is very small, which can affect the robustness of the model in the constraint environment. As a result, this paper proposes a sewer condition monitoring framework based on deep learning, which can effectively detect and evaluate defects in sewer pipelines with high accuracy. We also introduce a large dataset of sewer defects with 20 different defect classes found in the sewer pipeline. This study modified the original RegNet model by modifying the squeeze excitation (SE) block and adding the dropout layer and Leaky Rectified Linear Units (LeakyReLU) activation function in the Block structure of RegNet model. This study explored different deep learning methods such as RegNet, ResNet50, very deep convolutional networks (VGG), and GoogleNet to train on the sewer defect dataset. The experimental results indicate that the proposed system framework based on the modified-RegNet (RegNet+) model achieves the highest accuracy of 99.5 compared with the commonly used deep learning models. The proposed model provides a robust deep learning model that can effectively classify 20 different sewer defects and be utilized in real-world sewer condition monitoring applications.

Automated Identification of Overheated Belt Conveyor Idlers in Thermal Images with Complex Backgrounds Using Binary Classification with CNN.

Mechanical industrial infrastructures in mining sites must be monitored regularly. Conveyor systems are mechanical systems that are commonly used for safe and efficient transportation of bulk goods in mines. Regular inspection of conveyor systems is a challenging task for mining enterprises, as conveyor systems' lengths can reach tens of kilometers, where several thousand idlers need to be monitored. Considering the harsh environmental conditions that can affect human health, manual inspection of conveyor systems can be extremely difficult. Hence, the authors proposed an automatic robotics-based inspection for condition monitoring of belt conveyor idlers using infrared images, instead of vibrations and acoustic signals that are commonly used for condition monitoring applications. The first step in the whole process is to segment the overheated idlers from the complex background. However, classical image segmentation techniques do not always deliver accurate results in the detection of target in infrared images with complex backgrounds. For improving the quality of captured infrared images, preprocessing stages are introduced. Afterward, an anomaly detection method based on an outlier detection technique is applied to the preprocessed image for the segmentation of hotspots. Due to the presence of different thermal sources in mining sites that can be captured and wrongly identified as overheated idlers, in this research, we address the overheated idler detection process as an image binary classification task. For this reason, a Convolutional Neural Network (CNN) was used for the binary classification of the segmented thermal images. The accuracy of the proposed condition monitoring technique was compared with our previous research. The metrics for the previous methodology reach a precision of 0.4590 and an F1 score of 0.6292. The metrics for the proposed method reach a precision of 0.9740 and an F1 score of 0.9782. The proposed classification method considerably improved our previous results in terms of the true identification of overheated idlers in the presence of complex backgrounds.

Analyzing the Internal Part Condition on the Planetary Gear unit with Conditioning Monitoring Method at Cement Factory Horizontal Milling Station

Predictive maintenance is currently preferred over reactive maintenance because it’s very effective in reducing the costs incurred for unplanned breakdown stalling of a machine or equipment while production is running. In a production process at a Cement factory, the Milling station process is very important and becomes the main process. So, if something goes wrong in this process, the production line can be stopped, and the losses due to this are very big. Condition monitoring is a significant part of predictive maintenance because the application of condition monitoring allows the scheduled maintenance and precautions to be taken to prevent further failures and possible unplanned downtimes at a later date. By measuring the value of the vibration level, we can know whether the bearings and gear teeth on the machine are suitable for use or not. Based on this analysis, we can conclude that there is no sign of defect both on bearing in each gear teeth and the gear teeth inside the gearbox.

Application of Condition Monitoring for Hydraulic Oil Using Tuning Fork Sensor: A Study Case on Hydraulic System of Earth Moving Machinery.

In this study, we focus on the correctness of oil condition monitoring, specifically of a tuning forks sensor in hydraulic systems. We also aim to analyze the correlation between the online monitoring sensor signal and offline oil analysis by periodically sampling the hydraulic oil. In recent years, condition-based monitoring (CBM) of hydraulic oils has played a key role in extending earthmoving machinery uptime and reducing maintenance costs. We performed rig test and field test to develop a condition monitoring system based on oil analysis for construction equipment. Using the rig test, a reference line for the diagnosis of viscosity and dielectric constant for the new hydraulic oil was derived, and the characteristics of each sensor parameter for artificial contamination and oxidation were confirmed. In order to affirm the validity of oil diagnosis using oil sensors, the oil sensors were applied to four excavators to detect changes in oil conditions over 12 months. It was found that monitoring hydraulic oil with an oil sensor detecting the change in oil properties and contamination can provide reliable information for establishing diagnostic criteria. The finding allows us to predict the remaining oil life and to determine the oil change intervals based on the diagnosis of the oil condition.

A Predictive Maintenance System Design and Implementation for Intelligent Manufacturing

The importance of predictive maintenance (PdM) programs has been recognized across many industries. Seamless integration of the PdM program into today’s manufacturing execution systems requires a scalable and generic system design and a set of key performance indicators (KPIs) to make condition monitoring and PdM activities more effective. In this study, a new PdM system and its implementation are presented. KPIs and metrics are proposed and implemented during the design to enhance the system and the PdM performance monitoring needs. The proposed system has been tested in two independent use cases (autonomous transfer vehicle and electric motor) for condition monitoring applications to detect incipient equipment faults or operational anomalies. Machine learning-based data augmentation tools and models are introduced and automated with state-of-the-art AutoML and workflow automation technologies to increase the system’s data collection and data-driven fault classification performance.

Condition Monitoring of DC-Link Capacitors Using Time–Frequency Analysis and Machine Learning Classification of Conducted EMI

Condition monitoring techniques for power electronics components are important for reducing maintenance costs and increasing reliability in systems such as aircraft. This article presents a noninvasive condition monitoring system that utilizes time-frequency analysis of conducted electromagnetic interference (EMI) to classify the health of the dc-link capacitor within a three-phase inverter. The approach proposes a combined EMI filter and measurement board which is placed on the dc bus of the inverter. This board filters conducted EMI effectively and enables the inverter to comply with MIL-STD-461 G. It also enables EMI measurements to be collected for condition monitoring applications. The EMI content obtained from this board is analyzed from 15–43 MHz during switching events using a continuous wavelet transform. These characteristic switching images are used to train support vector machine models that are able to classify dc-link health into one of five health stages with accuracy up to 100%.

Anomaly detection-based condition monitoring

The impact of an anomaly is domain-dependent. In a dataset of network activities, an anomaly can imply an intrusion attack. Other objectives of anomaly detection are industrial damage detection, data leak prevention, identifying security vulnerabilities or military surveillance. Anomalies are observations or a sequence of observations in which the distribution deviates remarkably from the general distribution of the whole dataset. A large majority of the dataset consists of normal (healthy) data points. The anomalies form only a very small part of the dataset. Anomaly detection is the technique used to find these observations and its methods are specific to the type of data. While there is a wide spectrum of anomaly detection approaches available today, it becomes increasingly difficult to keep track of all the techniques. In fact, it is not clear which of the three categories of detection methods, ie statistical approaches, machine learning approaches or deep learning approaches, is more appropriate to detect anomalies in time-series data, which are mainly used in industry. A typical industrial device has multi-dimensional characteristics. It is possible to measure voltage, current, active power, vibrations, rotational speed, temperature, pressure difference, etc, on such a device. Early detection of the anomalous behaviour of industrial devices can help reduce or prevent serious damage, which could lead to significant financial loss. This paper presents a summary of the methods used to detect anomalies in condition monitoring applications.

Recognition of fatigue and severe sliding wear particles using a CNN model with multi-scale feature extractor

PurposeThe purpose of this study is to provide a new convolutional neural network (CNN) model with multi-scale feature extractor to segment and recognize wear particles in complex ferrograph images, especially fatigue and severe sliding wear particles, which are similar in morphology while different in wear mechanism.Design/methodology/approachA CNN model named DWear is proposed to semantically segment fatigue, severe sliding particles and four other types of particles, that is, chain, spherical, cutting and oxide particles, which unifies segmentation and recognition together. DWear is constructed using four modules, namely, encoder, densely connected atrous spatial pyramid pooling, decoder and fully connected conditional random field. Different from the architectures of ordinary semantic segmentation CNN models, a multi-scale feature extractor using cascade connections and a coprime atrous rate group is incorporated into the DWear model to obtain multi-scale receptive fields and better extract features of wear particles. Moreover, fully connected conditional random field module is adopted for post-processing to smooth coarse prediction and obtain finer results.FindingsDWear is trained and verified on the ferrograph image data set, and experimental results show that the final Mean Pixel Accuracy is 95.6% and the Mean Intersection over Union is 92.2%, which means that the recognition and segmentation accuracy is higher than those of previous works.Originality/valueDWear provides a promising approach for wear particle analysis and can be further developed in equipment condition monitoring applications.

SOTIF risk mitigation based on unified ODD monitoring for autonomous vehicles

Purpose The purpose of this paper is to design a unified operational design domain (ODD) monitoring framework for mitigating Safety of the Intended Functionality (SOTIF) risks triggered by vehicles exceeding ODD boundaries in complex traffic scenarios. Design/methodology/approach A unified model of ODD monitoring is constructed, which consists of three modules: weather condition monitoring for unusual weather conditions, such as rain, snow and fog; vehicle behavior monitoring for abnormal vehicle behavior, such as traffic rule violations; and road condition monitoring for abnormal road conditions, such as road defects, unexpected obstacles and slippery roads. Additionally, the applications of the proposed unified ODD monitoring framework are demonstrated. The practicability and effectiveness of the proposed unified ODD monitoring framework for mitigating SOTIF risk are verified in the applications. Findings First, the application of weather condition monitoring demonstrates that the autonomous vehicle can make a safe decision based on the performance degradation of Lidar on rainy days using the proposed monitoring framework. Second, the application of vehicle behavior monitoring demonstrates that the autonomous vehicle can properly adhere to traffic rules using the proposed monitoring framework. Third, the application of road condition monitoring demonstrates that the proposed unified ODD monitoring framework enables the ego vehicle to successfully monitor and avoid road defects. Originality/value The value of this paper is that the proposed unified ODD monitoring framework establishes a new foundation for monitoring and mitigating SOTIF risks in complex traffic environments.

Scalable change-point and anomaly detection in cross-correlated data with an application to condition monitoring

Motivated by a condition monitoring application arising from subsea engineering, we derive a novel, scalable approach to detecting anomalous mean structure in a subset of correlated multivariate time series. Given the need to analyse such series efficiently, we explore a computationally efficient approximation of the maximum likelihood solution to the resulting modelling framework and develop a new dynamic programming algorithm for solving the resulting binary quadratic programme when the precision matrix of the time series at any given time point is banded. Through a comprehensive simulation study we show that the resulting methods perform favorably compared to competing methods, both in the anomaly and change detection settings, even when the sparsity structure of the precision matrix estimate is misspecified. We also demonstrate its ability to correctly detect faulty time periods of a pump within the motivating application.