Industrial Condition Monitoring Research Articles

A semantic framework for condition monitoring in Industry 4.0 based on evolving knowledge bases

In Industry 4.0, factory assets and machines are equipped with sensors that collect data for effective condition monitoring. This is a difficult task since it requires the integration and processing of heterogeneous data from different sources, with different temporal resolutions and underlying meanings. Ontologies have emerged as a pertinent method to deal with data integration and to represent manufacturing knowledge in a machine-interpretable way through the construction of semantic models. Ontologies are used to structure knowledge in knowledge bases, which also contain instances and information about these data. Thus, a knowledge base provides a sort of virtual representation of the different elements involved in a manufacturing process. Moreover, the monitoring of industrial processes depends on the dynamic context of their execution. Under these circumstances, the semantic model must provide a way to represent this evolution in order to represent in which situation(s) a resource is in during the execution of its tasks to support decision making. This paper proposes a semantic framework to address the evolution of knowledge bases for condition monitoring in Industry 4.0. To this end, firstly we propose a semantic model (the COInd4 ontology) for the manufacturing domain that represents the resources and processes that are part of a factory, with special emphasis on the context of these resources and processes. Relevant situations that combine sensor observations with domain knowledge are also represented in the model. Secondly, an approach that uses stream reasoning to detect these situations that lead to potential failures is introduced. This approach enriches data collected from sensors with contextual information using the proposed semantic model. The use of stream reasoning facilitates the integration of data from different data sources, different temporal resolutions as well as the processing of these data in real time. This allows to derive high-level situations from lower-level context and sensor information. Detecting situations can trigger actions to adapt the process behavior, and in turn, this change in behavior can lead to the generation of new contexts leading to new situations. These situations can have different levels of severity, and can be nested in different ways. Dealing with the rich relations among situations requires an efficient approach to organize them. Therefore, we propose a method to build a lattice, ordering those situations depending on the constraints they rely on. This lattice represents a road-map of all the situations that can be reached from a given one, normal or abnormal. This helps in decision support, by allowing the identification of the actions that can be taken to correct the abnormality avoiding in this way the interruption of the manufacturing processes. Finally, an industrial application scenario for the proposed approach is described.

Structured Data Ontology for AI in Industrial Asset Condition Monitoring

This paper proposes an ontology for prognostics and health management (PHM) applications involving sensor networks monitoring industrial machinery. Deep learning methods show promise for the development of autonomous PHM systems but require vast quantities of structured and representative data to realize their potential. PHM systems involve unique and specialized data characterized by time and context, and thus benefit from tailored data management systems. Furthermore, the use of dissimilar standards and practices with respect to database structure and data organization is a hinderance to interoperability. To address this, this paper presents a robust, structured data ontology and schema that is designed to accommodate a wide breadth of PHM applications. The inclusion of contextual and temporal data increases its value for developing and deploying enhanced ML-driven PHM systems. Challenges around balancing the competing priorities of structure and flexibility are discussed. The proposed schema provides the benefits of a relational schema with some provisions for noSQL-like flexibility in areas where PMH applications demand it. The selection of a database engine for implementation is also discussed, and the proposed ontology is demonstrated using a Postgres database. An instance of the database was loaded with large auto-generated fictitious data via multiple Python scripts. CRUD (create, read, update, delete) operations are demonstrated with several queries that answer common PHM questions.

Labelling of Annotated Condition Monitoring Data Through Technical Language Processing

We propose a novel approach to facilitate supervised fault diagnosis on unlabelled but annotated industry datasets using human-centric technical language processing and weak supervision. Fault diagnosis through Condition Monitoring (CM) is vital for high safety and resource efficiency in the green transition and digital transformation of the process industry. Learning-based Intelligent Fault Diagnosis (IFD) methods are required to automate maintenance decisions and improve decision support for analysts. A major challenge is the lack of labelled industry datasets, limiting supervised IFD research to lab datasets. However, features learned from lab environments generalise poorly to field environments due to different signal distributions, artificial induction or acceleration of lab faults, and lab set-up properties such as average frequency profiles affecting learned features. In this study, we investigate how the unstructured free text fault annotations and maintenance work orders that are present in many industrial CM systems can be used for IFD through technical language processing, based on recent advances in natural language supervision. We introduce two distinct pipelines, one based on contrastive pre-training on large datasets, and one based on a small-data human-centric approach with unsupervised clustering methods. Finally, we showcase one example of the small-data fault classification implementation on a CM industry dataset with a SentenceBERT language model, kMeans clustering, and conventional signal processing methods. Fault class imbalance and time-shift uncertainty is overcome with weak supervision through aggregates of features, and human-centric clustering is used to integrate technical knowledge with the annotation-based fault classes. We show that our model can separate cable and sensor fault recordings from bearing-related fault recordings with an F1-score of 93. To our knowledge, this is the first system to classify faults in field industry CM data based only on associated unstructured fault annotations.

Statistical process control versus deep learning for power plant condition monitoring

This study compares four models for industrial condition monitoring including a principal components analysis (PCA) approach and three deep learning models, one of which is a new, lightweight version of another. We also propose a simple attention mechanism for enchancing deep learning models with better predictions and feature importance. Two datasets are used, one simulated from the Tennessee Eastman Process, the other from two feedwater pumps at a Danish combined heat and power plant. Our final results show evidence in favour of the PCA-based approach as it has detection ability comparable to the deep learning approaches as well as faster training time, fewer hyperparameters, as well as robustness to changing operating conditions. We conclude the paper by putting into perspective the importance of building up complexity incrementally with a recommendation to start modelling with simpler and well-tested models before the adoption of more advanced, less transparent models.

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.

Unsupervised Extraction of Shape-Based Signal Patterns for Incoming Signal Recognition

Coming up with a method to extract meaningful information from sensor signals without prior knowledge is one of the important challenges in machine learning. In the context of this challenge, this paper develops an unsupervised method to extract high-intensity shape-based signal patterns collected by a spintronic sensor for industrial machine condition monitoring. For this purpose, this paper first proposes a feature extraction method inspired by both shapelet transform and wavelet transform. The feature extraction method first extracts unsupervised shapelets, which are then treated as mother wavelets or filters to decompose the signals into their components. In addition, an idea based on the signal symbolic representation is proposed in order to extract signal patterns of various lengths. The signals are summarized with symbols obtained by finding a high-quality clustering result using an internal clustering validation criterion, which makes it suitable for further data mining tasks using far fewer memory resources. Finally, the signal patterns or symbols extracted from recorded data are then used for real-time signal recognition by a measure proposed to compute the similarity of the incoming signal with all the existing symbols. The performance and efficiency of our approach are demonstrated using real datasets.

Global Mechanical Response Sensing of Corrugated Compensators Based on Digital Twins

The corrugated compensators are important components in the piping system, absorbing mechanical deformation flexibly. To reduce the risk of the piping system with corrugated compensators and improve the safety and stability of industrial equipment, condition monitoring and fault diagnosis of bellows is necessary. However, the stress monitoring method of corrugated compensators with limited localized sensors lack real-time and full-domain sensing. Therefore, this paper proposes a digital twin construction method for global mechanical response sensing of corrugated compensators, combining Gaussian process regression in machine learning and finite element analysis. The sensing data of three types of displacements are used as the associated information of a finite element model with 19,800 elements and its digital twin. The results show that the values of performance metrics correlation of determination R2 and standardized average leave-one-out cross-validation CVavg of the digital twin satisfy the recommended threshold, which indicates that the digital twin has excellent predictive performance. The single prediction time of the digital twin is 0.76% of the time spent on finite element analysis, and the prediction result has good consistency with the true response under dynamic input, indicating that the digital twin can achieve fast and accurate stress field prediction. The important state information hidden in the multi-source data obtained by limited sensors is effectively mined to achieve the real-time prediction of the stress field. This paper provides a new approach for intelligent sensing and feedback of corrugated compensators in the piping system.

Explainable AI for Machine Fault Diagnosis: Understanding Features’ Contribution in Machine Learning Models for Industrial Condition Monitoring

Although the effectiveness of machine learning (ML) for machine diagnosis has been widely established, the interpretation of the diagnosis outcomes is still an open issue. Machine learning models behave as black boxes; therefore, the contribution given by each of the selected features to the diagnosis is not transparent to the user. This work is aimed at investigating the capabilities of the SHapley Additive exPlanation (SHAP) to identify the most important features for fault detection and classification in condition monitoring programs for rotating machinery. The authors analyse the case of medium-sized bearings of industrial interest. Namely, vibration data were collected for different health states from the test rig for industrial bearings available at the Mechanical Engineering Laboratory of Politecnico di Torino. The Support Vector Machine (SVM) and k-Nearest Neighbour (kNN) diagnosis models are explained by means of the SHAP. Accuracies higher than 98.5% are achieved for both the models using the SHAP as a criterion for feature selection. It is found that the skewness and the shape factor of the vibration signal have the greatest impact on the models’ outcomes.

A Review of Real-Time Fault Diagnosis Methods for Industrial Smart Manufacturing

In the era of Industry 4.0, highly complex production equipment is becoming increasingly integrated and intelligent, posing new challenges for data-driven process monitoring and fault diagnosis. Technologies such as IIoT, CPS, and AI are seeing increasing use in modern industrial smart manufacturing. Cloud computing and big data storage greatly facilitate the processing and management of industrial information flow, which helps the development of real-time fault diagnosis (RTFD) technology. This paper provides a comprehensive review of the latest RTFD technologies in the field of industrial process monitoring and machine condition monitoring. The RTFD process is introduced in detail, starting with the data acquisition process. The current RTFD methods are divided into methods based on independent feature extraction, methods based on “end-to-end” neural networks, and methods based on qualitative knowledge reasoning from a new perspective. In addition, this paper discusses the challenges and potential trends of RTFD in future development to provide a reference for researchers focusing on this field.

Tool life prognostics in CNC turning of AISI 4140 steel using neural network based on computer vision

One of the essential requirements for intelligent manufacturing is the low cost and reliable predictions of the tool life during machining. It is crucial to monitor the condition of the cutting tool to achieve cost-effective and high-quality machining. Tool conditioning monitoring (TCM) is essential to determining the remaining useful tool life to assure uninterrupted machining to achieve intelligent manufacturing. The same can be done by direct and indirect tool wear measurement and prediction techniques. In indirect methods, the data is acquired from the sensors resulting in some ambiguity, such as noise, reliability, and complexity. However, in direct methods, the data is available in images resulting in significantly less chances of ambiguity with the proper data acquisition system. The direct methods, which provide higher accuracy than indirect methods, involve collecting images of worn tools at different stages of the machining process to predict the tool life. In this context, a novel tool wear prediction system is proposed to examine the progressive tool wear utilizing the artificial neural network (ANN). Experiments were performed on AISI 4140 steel material under dry cutting conditions with carbide inserts. The cutting speed, feed, depth of cut, and white pixel counts are considered as input parameters for the proposed model, and the flank wear along with remaining tool life is predicted as the output. The worn tool images were captured using an industrial camera during the turning operation at regular intervals. The ANN training set predicts the remaining useful tool life, especially the sigmoid function and rectified linear unit (ReLU) activation function of ANN. The sigmoid function showed an accuracy of 86.5%, and the ReLU function resulted in 93.3% accuracy in predicting tool life. The proposed model’s maximum and minimum root mean square error (RMSE) is 1.437 and 0.871 min. The outcomes showcased the ability of image processing and ANN modeling as the potential approach for developing a low-cost industrial tool condition monitoring system that can measure tool wear and predict tool life in turning operations.

Generation of reproducible and scalable local damage to rolling bearings for the development of condition monitoring systems through the use of laser technology

Abstract In the context of the fourth industrial revolution, condition monitoring and predictive maintenance of machines are becoming increasingly important, as unplanned machine failures can lead to high costs for the operator. Many of these machine failures can be traced back to bearing damage, which must be prevented. In order to be able to make reliable statements about the remaining service life of the bearings, a condition monitoring system is required. For the development of these systems, bearings with a defined level of damage are needed. Since there are not enough reproducible methods for bearing damage so far, the defined generation of reproducible, local bearing damage is researched by using laser technology. The damage caused is checked and evaluated by means of vibration analysis and white light interferometry. The results confirm a reproducibility and scalability of the bearing damages by means of laser technology. The damaged bearings can be used for research-based development as well as in industrial implementations of condition monitoring systems.

Machinery condition monitoring in the era of industry 4.0: A relative degree of contribution feature selection and deep residual network combined approach

Condition monitoring (CM) is gaining importance under the profound and comprehensive impact of Industry 4.0 on the modern manufacturing industry, which can monitor the health status of machinery, and support its improvement on availability and reliability. With industrial machinery system tends to be increasingly complex and large, its CM signal processing often suffers from large data scale, redundant information, and mismatched sampling rate among heterogeneous sensors. Besides, though deep learning technology exhibits outstanding ability in data analysis, its application in machinery operational status analysis faces challenges posed by the CM sensing data encoding problem, which aims to match the data entry conditions of the deep learning model, and the deep network efficient training problem. In this paper, we present an industrial machinery condition monitoring analysis method combining the relative degree of contribution (RDoC) based feature selection and deep residual network (DRN). A novel extendable RDoC based feature selection strategy is proposed to identify the optimal feature combination with high representational information density. Information contribution of both sensors and signal features are considered respectively in the feature selection process. Then after a delicately designed matrix-encoding operation, the extracted feature combination is regarded as the input of DRN, to support operational status pattern analysis for machinery CM. Experiment studies on a real dataset demonstrate the feasibility and efficiency of this method. The research will provide a valuable reference to achieve effective monitoring of machinery in Industry 4.0 era.

Comparison of different ML methods concerning prediction quality, domain adaptation and robustness

Abstract Nowadays machine learning methods and data-driven models have been used widely in different fields including computer vision, biomedicine, and condition monitoring. However, these models show performance degradation when meeting real-life situations. Domain or dataset shift or out-of-distribution (OOD) prediction is mentioned as the reason for this problem. Especially in industrial condition monitoring, it is not clear when we should be concerned about domain shift and which methods are more robust against this problem. In this paper prediction results are compared for a conventional machine learning workflow based on feature extraction, selection, and classification/regression (FESC/R) and deep neural networks on two publicly available industrial datasets. We show that it is possible to visualize the possible shift in domain using feature extraction and principal component analysis. Also, experimental competition shows that the cross-domain validated results of FESC/R are comparable to the reported state-of-the-art methods. Finally, we show that the results for simple randomly selected validation sets do not correctly represent the model performance in real-world applications.

Towards Interpretable Machine Learning for Automated Damage Detection Based on Ultrasonic Guided Waves.

Data-driven analysis for damage assessment has a large potential in structural health monitoring (SHM) systems, where sensors are permanently attached to the structure, enabling continuous and frequent measurements. In this contribution, we propose a machine learning (ML) approach for automated damage detection, based on an ML toolbox for industrial condition monitoring. The toolbox combines multiple complementary algorithms for feature extraction and selection and automatically chooses the best combination of methods for the dataset at hand. Here, this toolbox is applied to a guided wave-based SHM dataset for varying temperatures and damage locations, which is freely available on the Open Guided Waves platform. A classification rate of 96.2% is achieved, demonstrating reliable and automated damage detection. Moreover, the ability of the ML model to identify a damaged structure at untrained damage locations and temperatures is demonstrated.

Fault Detection and Identification Using Deep Learning Algorithms in Induction Motors

Owing to the 4.0 industrial revolution condition monitoring maintenance is widely accepted as a useful approach to avoiding plant disturbances and shutdown. Recently, Motor Current Signature Analysis (MCSA) is widely reported as a condition monitoring technique in the detection and identification of individual and multiple Induction Motor (IM) faults. However, checking the fault detection and classification with deep learning models and its comparison among themselves or conventional approaches is rarely reported in the literature. Therefore, in this work, we present the detection and identification of induction motor faults with MCSA and three Deep Learning (DL) models namely MLP, LSTM, and 1D-CNN. Initially, we have developed the model of Squirrel Cage induction motor in MATLAB and simulated it for single phasing and stator winding faults (SWF) using Fast Fourier Transform (FFT), Short Time Fourier Transform (STFT), and Continuous Wavelet Transform (CWT) to detect and identify the healthy and unhealthy conditions with phase to ground, single phasing and in multiple fault conditions using Motor Current Signature Analysis. The faults impact on stator current is presented in the time and frequency domain (i.e., power spectrum). The simulation results show that the scalogram has shown good results in time-frequency analysis for fault and showing its impact on the energy of current during individual fault and multiple fault conditions. This is further investigated with three deep learning models (i.e., MLP, LSTM, and 1D-CNN) for checking the fault detection and identification (i.e., classification) improvement in a three-phase induction motor. By simulating the three-phase induction motor in various healthy and unhealthy conditions in MATLAB, we have collected current signature data in the time domain, labeled them accordingly and created the 50 thousand samples dataset for DL models. All the DL models are trained and validated with a suitable number of architecture layers. By simulation, the multiclass confusion matrix, precision, recall, and F1-score are obtained in several conditions. The result shows that the stator current signature of the motor can be used to detect individual and multiple faults. Moreover, deep learning models can efficiently classify the induction motor faults based on time-domain data of the stator current signature. In deep learning (DL) models, the LSTM has shown better accuracy among all other three models. These results show that employing deep learning in fault detection and identification of induction motors can be very useful in predictive maintenance to avoid shutdown and production cycle stoppage in the industry.

Physically Flexible Ultralow-Power Wireless Sensor

The key challenges of local sensor networks are in supporting high sensor density, information security, physical size, and especially energy efficiency at a level that could eliminate the need for batteries or external power supplies. This article presents a novel scheme that answers all issues at the cost of minor information losses in low data rate applications that tolerate latency. Experimental verification is made using a sensor node implemented on a flexible electronics platform. Lightly encrypted data is transmitted by embedding it into Bluetooth advertising packets, contributing to ultra-low-energy wireless power consumption, and theoretically enabling an unlimited number of nodes in the local network. In the experiments the energy dissipation per transmitted 14 bytes information packet varied between 19.83 μW and 105.93 μW depending on the system configuration, while the data loss rates ranged from 7.4 % to 0.004 %, respectively. As the flexible substrate can be attached to various surfaces, the applications extend from wearable to industrial condition monitoring devices.

Physics-informed deep learning for signal compression and reconstruction of big data in industrial condition monitoring

The onset of the Internet of Things enables machines to be outfitted with always-on sensors that can provide health information to cloud-based monitoring systems for prognostics and health management (PHM), which greatly improves reliability and avoids downtime of machines and processes on the shop floor. On the other hand, real-time monitoring produces large amounts of data, leading to significant challenges for efficient and effective data transmission (from the shop floor to the cloud) and analysis (in the cloud). Restricted by industrial hardware capability, especially Internet bandwidth, most solutions approach data transmission from the perspective of data compression (before transmission, at local computing devices) coupled with data reconstruction (after transmission, in the cloud). However, existing data compression techniques may not adapt to domain-specific characteristics of data, and hence have limitations in addressing high compression ratios where full restoration of signal details is important for revealing machine conditions. This study integrates Deep Convolutional Autoencoders (DCAE) with local structure and physics-informed loss terms that incorporate PHM domain knowledge such as the importance of frequency content for machine fault diagnosis. Furthermore, Fault Division Autoencoder Multiplexing (FDAM) is proposed to mitigate the negative effects of multiple disjoint operating conditions on reconstruction fidelity. The proposed methods are evaluated on two case studies, and autocorrelation-based noise analysis provides insight into the relative performance across machine health and operating conditions. Results indicate that physically-informed DCAE compression outperforms prevalent data compression approaches, such as compressed sensing, Principal Component Analysis (PCA), Discrete Cosine Transform (DCT), and DCAE with a standard loss function. FDAM can further improve the data reconstruction quality for certain machine conditions.

Intelligent Diagnosis of Equipment Health Based on IOT and Operation Large Data Analysis

Predictive maintenance integrates equipment condition monitoring, fault diagnosis, fault prediction, maintenance decision support and maintenance activities. Intelligent manufacturing upgrades need to match the synchronous improvement of predictive maintenance capabilities, and predictive maintenance is the basic guarantee for enterprises to achieve intelligent manufacturing. Take the equipment condition monitoring and diagnosis system applicate in process industry as an example, this paper proposes IOT and operation big data analysis to equipment fault monitoring, diagnosis and preventive maintenance. Under the three-layer system framework of perception layer, network layer and application layer, machine learning algorithm is applied to carry out data mining on the equipment operation big data, establish expert knowledge base, obtain diagnosis rules, and realize the intelligent and efficient management mode integrating online monitoring, remote monitoring, remote diagnosis, fault matching and identification. Based on IOT and operation large data analysis, equipment health intelligent diagnosis provides the basic guarantee for equipment intelligent operation and maintenance, which helping enterprise establish new equipment management, maintenance, inspection and repair system under the concept of predictive maintenance and proactive maintenance.

A framework combining acoustic features extraction method and random forest algorithm for gas pipeline leak detection and classification

Monitoring the operation status of the gas pipeline is of great significance to ensure the safe and stable operation of the pipeline. A new framework combining the acoustic features extraction method and Random Forest (RF) algorithm is proposed for gas pipeline leak detection and classification under the strong background noise. Acoustic signal has the advantages of non-contact measurement and insensitive to structural resonances, but it is easy to collect signals with low Signal-to-Noise Ratio (SNR) under strong background noise. To improve SNR and extract useful acoustic features, the acoustic features extraction method based on Ensemble Empirical Mode Decomposition (EEMD) and Correlation Coefficient (CC) analysis is proposed. CC analysis is used to select appropriate Intrinsic Mode Functions (IMFs), which are decomposed by EEMD algorithm. For each selected IMF, Mel-Frequency Cepstral Coefficients (MFCC), time-domain features, waveform features are extracted to form a feature matrix. Because RF has the characteristic of a small training set requirement, which provides a solution to the difficulty of obtaining enough leak acoustic signals in the industrial environment, RF algorithm is employed for leak detection and classification. Experiments were conducted to verify the effectiveness of the proposed framework. When the sampling distance is set to 1 m and 6 m, the average accuracies of RF with extracted acoustic features are higher than the results with all features. The proposed framework is general and can be applied to various acoustic-based industrial equipment condition monitoring challenges.

Enabling predictive maintenance integrated production scheduling by operation-specific health prognostics with generative deep learning

Predictive Maintenance (PdM) is one of the core innovations in recent years that sparks interest in both research and industry. While researchers develop more and more complex machine learning (ML) models to predict the remaining useful life (RUL), most models are not designed with regard to actual industrial practice and are not validated with industrial data. To overcome this gap between research and industry and to create added value, we propose a holistic framework that aims at directly integrating PdM models with production scheduling. To enable PdM-integrated production scheduling (PdM-IPS), an operation-specific health prognostics model is required. Therefore, we propose a generative deep learning model based on the conditional variational autoencoder (CVAE) that can derive an operation-specific health indicator (HI) from large-scale industrial condition monitoring (CM) data. We choose this unsupervised learning approach to cope with one of the biggest challenges of applying PdM in industry: the lack of labelled failure data. The health prognostics model provides a quantitative measure of degradation given a specific production sequence and thus enables PdM-IPS. The framework is validated both on NASA’s C-MAPSS data set as well as real industrial data from machining centers for automotive component manufacturing. The results indicate that the approach can both capture and quantify changes in machine condition such that PdM-IPS can be subsequently realized.