Robust Condition Monitoring Research Articles

Detection and classification of surface defects on hot-rolled steel using vision transformers

This study proposes a vision transformer to detect visual defects on steel surfaces. The proposed approach utilizes an open-source image dataset to classify steel surface conditions into six fault categories namely, crazing, inclusion, rolled in, pitted surface, scratches and patches. The defect images are first subject to resizing and then fed into a vision transformer subject to different hyperparameter configurations to determine the most optimal setting to render highest classification performance. The performance of the model is evaluated for different hyperparameter configurations, and the most optimal configuration is examined using the associated confusion matrices. It was observed that the proposed model presents a high overall accuracy of 96.39 % for detection and classification of steel surface faults. The study presents a descriptive insight into the vision transformer architecture and in addition, compares the performance of the current model with the results of other approaches suggested for application in literature. Vision transformers can serve as standalone approaches and suitable alternatives to the widely used convolution neural networks (CNNs) by actuating complex defect detection and classification tasks in real-time, enabling efficient and robust condition monitoring of a wide range of defects.

Vibration energy-based indicators for multi-target condition monitoring in milling operations

The demand for intelligent process monitoring is increasing in aerospace manufacturing to ensure tight tolerances and high surface quality. Real-time monitoring in machining is crucial for machined accuracy and process reliability, reducing production times and costs, and enhancing automation of the manufacturing process. This study presents a robust multi-target condition monitoring method based on the vibration signals. Firstly, three new energy ratio indicators with dimensionless characteristics were defined for tool wear, breakage, and chatter monitoring. Secondly, the vibration energy loss from the tool tip to the tool holder, and spindle housing was measured and compared, and the rules of vibration loss from the tool tip to the spindle housing were revealed. Using force signals as a reference, the monitoring performance of industrially acceptable acceleration and sound signals in multi-target condition monitoring was quantitatively analyzed. Finally, the performance of the proposed vibration energy-based indicators was experimentally illustrated and quantitatively evaluated. It is shown that these indicators can be used to discriminate between tool breakage and chatter, as well as to assess tool wear. The new monitoring method can also minimize the costs of process monitoring by reducing the use of expensive sensors or overusing multiple sensors in a smart manufacturing system.

A System Architecture for Early Wilt Detection in Hydroponic Crops: An Implementation and Assessment

Abstract Hydroponic farming systems have a high risk of failure if the disaster is not mitigated and responded to quickly. Hydroponic plants only rely on mineral water that flows through their roots as a source of life; thus, any malfunction will have an immediate impact on the plant. If plant wilting is not recognized and not treated quickly, the plant will experience stress and have the potential to fail. To ensure that major disasters do not occur, an early warning mechanism is needed that can provide disaster information to farmers. In this study, an intelligent system architecture was built to detect premature wilt in hydroponic vegetable plants. It combines the capabilities offered by the OV2640 image sensor and ESP32-S system-on-chip, the Internet of Things, deep learning based on convolutional neural networks, and cloud computing to create a robust and low-cost remote plant condition monitoring solution. As a proof-of-concept, a prototype system is built, and performance tests are carried out to find out how robust and effective the proposed design is. Based on the results of the prototype test, it can be seen that the system built can run properly with an accuracy of wilting recognition of up to 90.90%, which was tested on samples of vertically grown mustard plants in a hydroponic greenhouse.

Combination of generic novelty detection and supervised classification pipelines for industrial condition monitoring

Abstract Machine learning in industrial condition monitoring is currently a rapidly developing field of research, to improve the efficiency and reliability of industrial processes. Many of the used algorithms are supervised methods, which can learn and recognize hidden patterns in the data. However, training data is required to learn these patterns, which can only be generated to a limited extent in an industrial environment due to the high costs involved. Furthermore, it is impossible to represent all possible events in the training data. In contrast, unsupervised or semi-supervised methods can be used to detect new conditions or events. However, these usually do not allow diagnosis or quantification of a fault condition, which is why their usefulness for modern maintenance strategies is limited. Consequently, a robust condition monitoring system should combine the functionality of both approaches. This paper presents a methodology for the combination of supervised classification and semi-supervised novelty detection to build an expandable and adaptable condition monitoring by transferring recurring novelties as new conditions to the supervised classification. A superordinate algorithm is proposed to achieve a stepwise extension of the supervised model based on new conditions detected by novelty detection. With this approach, a condition monitoring system can at first be based on “normal” data of a new machine or process by adding failures or novel conditions step-by-step. Furthermore, the supervised methods can be used to help the corresponding staff identify unknown conditions by analyzing the features selected by the supervised classification. The general workflow is demonstrated for condition monitoring of the pneumatic drive system of a welding gun.

Hot rolled steel surface defect detection and classification using an automatic ensemble approach

This study introduces an ensemble-based Deep Neural Network (DNN) model for detecting defects on steel surfaces. The method suggested in this study classifies steel surface conditions into six possible fault categories, namely, crazing, inclusion, rolled in, pitted surface, scratches, and patches. The images undergo preprocessing and extraction of features in spatial and frequency domains using image segmentation techniques such as grey level difference method (GLDM), fast Fourier Transform (FFT), grey level co-occurrence matrix (GLCM), texture analysis and discrete wavelet transform (DWT). The ensembling of image features into a fused feature pool is carried out after the preprocessing of input images that are provided as input to a light-weight neural network model for training and testing. The performance of the model is comprehensively evaluated via an ablation study both before and after ensembling. In addition, the model capability is effectively analyzed using receiver operating characteristics (ROC) curve, confusion matrix from which classification accuracy of the model could be obtained and other parameters including precision and f1-score. It was observed that the proposed deep learning network presents phenomenally high accuracy of 99.72% for detection and classification of steel surface faults. This result was found to be superior when compared with the performance of the same neural network over each feature type individually. This study also compares the classification results of the model built based on the ensembled feature set with the results of various other classification approaches available in literature. The ensemble-based model could potentially be integrated into existing inspection systems for real-time, efficient and robust condition monitoring of steel surfaces.

Two-step vibration-based machine learning model for the fault detection and diagnosis in rotating machine and its blind application

A robust and reliable condition monitoring and fault diagnosis system is crucial for an efficient operation of industries. Because of the advances in technologies over the past few decades, there is an increased interest in developing intelligent systems to perform tasks that traditionally rely on knowledge, experience and expertise of an individual. It is known that unexpected breakdowns have wide implications in production processes. Thus, it is vital to be able to know the machine condition and detect at the earliest possible stage the defects when they occur. Aiming at an industrial application, in this study, a two-step approach is proposed for the fault detection and diagnosis of rotor-related faults. The implemented algorithm is a pattern recognition supervised artificial neural network, which through information extracted from vibration signals allows one to identify the health status of the machine. In the first step, the model identifies whether the machine is healthy or faulty. This is important information for any industry to operate the machines. Once the machine condition (healthy or faulty) is known and if it is faulty, then only faulty machine parameters are used in the second step to know the specific fault. The model is initially based on existing experimental data, and then, it is further validated with mathematically generated data. The proposed two-step approach model and the trained framework are applied blindly at a different machine speed, where the dynamics of machine is expected to be different. The excellent results obtained suggest this approach as a possibility for industrial application.

Predictive maintenance of rotational machinery using deep learning

This paper describes an implementation of a deep learning-based predictive maintenance (PdM) system for industrial rotational machinery, built upon the foundation of a long short-term memory (LSTM) autoencoder and regression analysis. The autoencoder identifies anomalous patterns, while the latter, based on the autoencoder’s output, estimates the machine’s remaining useful life (RUL). Unlike prior PdM systems dependent on labelled historical data, the developed system doesn’t require it as it’s based on an unsupervised deep learning model, enhancing its adaptability. The paper also explores a robust condition monitoring system that collects machine operational data, including vibration and current parameters, and transmits them to a database via a Bluetooth low energy (BLE) network. Additionally, the study demonstrates the integration of this PdM system within a web-based framework, promoting its adoption across various industrial settings. Tests confirm the system's ability to accurately identify faults, highlighting its potential to reduce unexpected downtime and enhance machinery reliability.

Fibre Bragg Grating Sensors for Condition Monitoring of High-Voltage Assets: A Review

The high-voltage (HV) assets in the existing power transmission network will experience increased electrical, thermal, environmental and mechanical stresses and, therefore, robust condition monitoring is critical for power system reliability planning. Fibre Bragg grating (FBG) sensors offer a promising technology in HV applications due to their immunity to electromagnetic interference and multiplexing capability. This paper reviews the current technology readiness levels of FBG sensors for condition monitoring of transformers, transmission lines, towers, overhead insulators and power cables, with the aim of stimulating further development and deployment of fibre-based HV asset management systems. Currently, there are several reported cases of FBG sensors used for condition monitoring of HV assets in the field, proving their feasibility for long-term use in the power grid. The review shows that FBG technology is versatile and can facilitate multi-parameter measurements, which will standardise the demodulation equipment and reduce challenges with integrating different sensing technologies.

Robust tool condition monitoring in Ti6Al4V milling based on specific force coefficients and growing self-organizing maps

Tool condition monitoring (TCM) is a mean to optimize production systems trying to use cutting tool life at its best. Nevertheless, nowadays available TCM algorithms typically lack robustness in order to be consistently applied in industrial scenarios. In this paper, an unsupervised artificial intelligence technique, based on Growing Self-Organizing Maps (GSOM), is presented in synergy with real-time specific force coefficients (SFC) estimation through the regression of instantaneous cutting forces. The conceived approach allows robustly mapping the SFC, exploiting process parameters and similarity to manage the variability of their estimation due to unmodelled phenomena, like machine dynamics and tool run-out. The devised approach allowed detecting the tool end-of-life in cutting tests with variable lubrication, machine tool and cutting speed, through the adoption of a self-starting control chart running on real-time clustered data. The solution was validated through the comparison of the GSOM framework with respect to the optimized self-starting control chart applied without GSOM clustering. The GSOM reached a root mean squared percentage error (RMSPE) of 13.2% with respect to 56.1% obtained with the analogous control chart in a full-set optimization scenario. When optimised on tests for a unique machine tool and tested on another machine tool, GSOM scored an RMSPE of 34.5%, whereas the optimized control chart scored 64.5%.

A Robust Tool Condition Monitoring System Based on Cluster Density under Variable Machining Processes

Real-time tool condition monitoring (TCM) is becoming more and more important to meet the increased requirement of reducing downtime and ensuring the machining quality of manufacturing systems. However, it is difficult to satisfy both robustness and effectiveness of pattern recognition for a TCM system without using an unsupervised strategy. In this paper, a clustering-based TCM system is proposed that can be used for different machining conditions such as variable cutting parameters, variable cutters, and even variable cutting methods. The solution is based on a significant statistical correlation between tool wear and the distribution of cutting force features, which is revealed through the clustering results obtained from a novel clustering method based on adjacent grids searching (CAGS). This statistical correlation is converted into tool wear status by using an empirical factor that is robust for variable cutting processes. The proposed TCM system is completely unsupervised as a training-free procedure is used in the monitoring process. To verify the effectiveness of the system, a series of experiments are conducted, such as whole life-cycle wear experiment under same milling condition, tool wear experiment under variable milling conditions and tool wear experiment under same turning condition. The prediction accuracy of our system for tool wear experiment under variable milling conditions is 100%, 75% and 75%, respectively. In contrast, BP neural network, Bayesian network and SVM are used for tool wear prediction under the same conditions. Experimental results show the superiority and effectiveness of our TCM system based on cluster density of CAGS over several state-of-the-art supervised methods.

The Influence of Gravitational Stiffening and De-stiffening on Nonlinear Dynamic Behavior of Vertical Beams with Cracks

ABSTRACT Dynamics of vertically oriented blades is an important topic of research as it replicates the physical behaviour of several real-time applications such as horizontal axis wind turbine blades. It is observed that available literatures explaining the effects of gravity on the dynamic behaviour of vertical beams/blades did not focus on the various operational adversities such as geometric nonlinearities, possibilities of crack initiation due to adverse conditions. Therefore, in present study, an attempt is made to capture reverberations of gravity on cracked vertical blades including geometric nonlinearity. A computationally efficient numerical model for nonlinear dynamic analysis of vertically oriented cracked blades is developed. Efficacy of proposed model is verified by several case studies. Subsequently, parametric investigation is performed for combined effects of crack and gravity on the dynamic response. Crack and gravity dominant regions based on slenderness ratio are identified for the first time . Furthermore, influence of nonlinearities on such behaviour provides useful insights to anticipate the mechanics of the dynamic behaviour of a vertical cracked cantilever beam. Proposed work may be useful for developing a robust vibration-based condition monitoring for a given application which involves hanging beams with a possibility of crack such as rotor blades.

State-of-the-Art Wayside Condition Monitoring Systems for Railway Wheels: A Comprehensive Review

In recent decades, there has been a constant demand for faster, longer, and safer railway networks. This also brings challenges for condition monitoring systems in modern railway vehicles. More specifically, critical parts of railway vehicles like wheels degrade over time due to various operational and environmental reasons. Different dynamic effects such as skidding/sliding over the track and the presence of contamination between wheel-rail cause various wheel defects. Faulty wheels ultimately lead to the derailment of railway vehicles. To avoid worst situations like railway derailments, various research has been conducted for developing efficient condition monitoring systems for railway wheels. In addition, there have been some commercial condition monitoring products that can be deployed with railway vehicles. These systems incorporate various sensors such as strain gauges and vision sensors to collect data for diagnosis and prognosis. Various methods have been explored but yet there is a broad research gap in terms of developing advanced onboard condition monitoring systems. With the progress in technology, advanced systems with Machine Learning/Deep Learning methods can provide more efficient and robust condition monitoring of dynamic railway systems. Considering the need for advancement in condition monitoring systems for railway vehicles, a comprehensive review of existing condition monitoring systems for railway wheels is conducted in this paper. The review is aimed at understanding the feasibility and potential of new methods for modern railways. This paper provides a detailed overview of studies on the existing wayside systems and reports their advantages and disadvantages concerning its recently emerging counterpart on-board monitoring systems. Data acquisition systems and analysis methods are critically reviewed which could assist in developing more efficient and reliable condition monitoring systems for railway wheels. This article also reviews the current progress of wayside systems and their limitations. The article is targeted at the researchers and engineers working in this domain, who can pave the way for developing advanced and cost-effective condition monitoring systems for railway wheels using modern technologies.

An improved first-principle model of AC powered solenoid operated valves for maintenance applications

Solenoid operated valves (SOVs) are critical components in many industrial applications. There has been a continuing interest in the industry to have robust condition monitoring, prognostics and health management tools to support the condition based maintenance and predictive maintenance program for such valves. For critical assets like SOVs, it is of paramount interest to understand why a component might be declared as defective. In such a situation, a first principle-model based approach will always be preferred to a purely data-driven approach, because of its inherent interpretability. Furthermore, first principle-models typically have less free parameters than their data driven counterparts and will require less data to identify their parameters. In this paper, we present the improvement of a first-principle model of alternating current (AC) powered SOVs taking into account two important degradation effects. Using this model, we show that the state of degradation can be estimated from current and input voltage measurement signals on the solenoids. Our method is validated using data from an accelerated life test campaign on 48 identical AC-powered SOVs.

Individual functionalization of fiber-reinforced profiles via pultrusion

Profile-based design is widely used because of its flexibility. A good possibility to increase the lightweight aspect can be the use of fiber reinforced profiles. These can be produced in a series suitable process - pultrusion. In this paper, the integration of functions in pultrusion profiles is focused. In one project, potential sensor concepts for robust condition monitoring of fiber reinforced plastic (FRP) profiles were developed to expand the range of applications significantly. One application could be the substitution of complex aluminum profiles by fiber reinforced profiles with the advantage like high lightweight potential and low thermal expansion. Another objective was the integration of sensors and light-emitting diodes (LED) directly into the pultrusion profile. The objective of a second research project was the integration, wiring and test of sensors into a curved roller ski profile to detect loads during training. Several types of sensors were investigated to evaluate the parameters accuracy, positioning, and wiring at the best economic efficiency. The focus of the research and development of both projects was the integration of functions into pultruded profiles made of FRP. Successfully integrated functionalizing elements include sensors for strain, lighting elements, and touch sensors as well as temperature and humidity modules. The objective of the research project “SmartSensSki” was the integration, wiring and test of sensors into a curved roller ski profile to detect loads during training. Several types of sensors were investigated to evaluate the parameters accuracy, positioning, and wiring at the best economic efficiency. The focus of the research and development of both projects was the integration of functions into pultruded profiles made of FRP. Successfully integrated functionalizing elements include sensors for strain, lighting elements, and touch sensors as well as temperature and humidity modules.

An approach to robust condition monitoring in industrial processes using pythagorean membership grades

In this paper, a robust approach to improve the performance of a condition monitoring process in industrial plants by using Pythagorean membership grades is presented. The FCM algorithm is modified by using Pythagorean fuzzy sets, to obtain a new variant of it called Pythagorean Fuzzy C-Means (PyFCM). In addition, a kernel version of PyFCM (KPyFCM) is obtained in order to achieve greater separability among classes, and reduce classification errors. The approach proposed is validated using experimental datasets and the Tennessee Eastman (TE) process benchmark. The results are compared with the results obtained with other algorithms that use standard and non-standard membership grades. The highest performance obtained by the approach proposed indicate its feasibility.

Solar panel hotspot localization and fault classification using deep learning approach

There has been an exponential increase in Photovoltaic energy over the last decade. The size and the complexity of photovoltaic solar power plants are increasing, and it requires advanced and robust condition monitoring systems for ensuring their reliability. To this aim, a novel method is addressed for fault detection in photovoltaic panels through processing of thermal images of solar panels captured by a thermographic camera. In this paper, two advanced convolutional neural network models are used wherein the task of the first model is to classify the type of fault affecting the panel and the task of the second model is to identify the region of interest of the faulty panel. Proposed approach uses F1 score as a metric to compare several classification models of which the ResNet-50 transfer learning model achieves the highest score of 85.37 %. Mean Average Precision is used as an evaluation metric for object detection models wherein the highest scoring model is Faster R-CNN with a score of 67 %. This paper puts forth an approach to facilitate early identification and fault localization in Solar Panels by minimizing the amount of manual labour involved in the process.

The Deterministic Nature of Sensor-Based Information for Condition Monitoring of the Cutting Process

Condition monitoring of the cutting process is a core function of autonomous machining and its success strongly relies on sensed data. Despite the enormous amount of research conducted so far into condition monitoring of the cutting process, there are still limitations given the complexity underlining tool wear; hence, a clearer understanding of sensed data and its dynamical behavior is fundamental to sustain the development of more robust condition monitoring systems. The dependence of these systems on acquired data is critical and determines the success of such systems. In this study, data is acquired from an experimental setup using some of the commonly used sensors for condition monitoring, reproducing realistic cutting operations, and then analyzed upon their deterministic nature using different techniques, such as the Lyapunov exponent, mutual information, attractor dimension, and recurrence plots. The overall results demonstrate the existence of low dimensional chaos in both new and worn tools, defining a deterministic nature of cutting dynamics and, hence, broadening the available approaches to tool wear monitoring based on the theory of chaos. In addition, recurrence plots depict a clear relationship to tool condition and may be quantified considering a two-dimensional structural measure, such as the semivariance. This exploratory study unveils the potential of non-linear dynamics indicators in validating information strength potentiating other uses and applications.

Robust vibration-based faults diagnosis machine learning model for rotating machines to enhance plant reliability

Plant availability and reliability can be improved through a robust condition monitoring and fault diagnosis model to predict the current status (healthy or faulty) of any machines and critical assets. The model can then predict the exact fault for the faulty asset so that remedial maintenance can be carried out in a planned plant outage. Nowadays, the artificial intelligence (AI)-based machine learning (ML) model seems to be current trend to meet these requirements. Hence, the paper is also proposing such vibration-based faults diagnosis ML model through an experimental rotating rig. Here, the 2-Steps approach is used with the ML model to easy the industrial operation and maintenance process. The Step-1 provides the information about the asset health status such as healthy or faulty. The Step-2 then identifies the exact nature of fault to aid the decision making for the fault rectification and maintenance activities to avoid the risk of failure and enhance the reliability.

Defining acoustic emission-based condition monitoring indicators for monitoring piston rod seal and bearing wear in hydraulic cylinders

Fluid leakage from hydraulic cylinders is a major concern for the offshore industries as it directly affects hydraulic cylinder energy efficiency and causes environmental contamination. There have been attempts made in literature to develop robust condition monitoring techniques for hydraulic cylinders. However, most of these studies were performed to identify degradation of single components. Therefore, in this study, the aim is to monitor degradation of multiple components simultaneously in hydraulic cylinders using acoustic emissions. Experiments performed consist of three test phases and were performed using a hydraulic test rig. In the first test phase, the study is performed to identify acoustic emission features that can be used to monitor piston rod seal wear. In the second test phase, acoustic emission features are identified that can be used to understand bearing wear when unworn, semi-worn or worn piston rod seals are used in hydraulic test rig. In the third test phase, a run-to-failure test is conducted to identify acoustic emission features that can indicate fluid leakage initiation due to piston rod seal wear. The median frequency feature showed good repeatability in all the three test phases to identify piston rod seal wear, bearing wear and fluid leakage initiation during the initial stages in the hydraulic test rig. The proposed acoustic emission-based condition monitoring technique is robust and can be used for the hydraulic cylinders in the industries, as it identifies acoustic emission features based on particular frequency bands associated to specific components, making it less susceptible to noise from other components.

A robust condition monitoring methodology for grinding wheel wear identification using Hilbert Huang transform

Grinding is a finishing operation performed to obtain the desired finish on the component. Wheel wear is one of the primary constraints in achieving the desired productivity in grinding. A new methodology is proposed for accurate and timely identification of wheel wear in cylindrical grinding using Hilbert Huang transform and support vector machine. During the grinding of EN31 carbon steel, the condition of the wheel and its wear was monitored with an accelerometer and power cell. Both vibration and power signals captured were used to identify the condition of the wheel and its wear. An exhaustive feature set is generated in the frequency and the time-frequency domain. Hilbert Huang transform, an adaptive time-frequency analysis technique, was used to extract the features of tool wear in the time-frequency domain. The first three IMF constituents were further chosen for feature extraction of statistical parameters based on their mean energy. Random forests algorithm was used to identify the relevant features. The methodology was validated with several grinding experiments and, is found to give an accuracy of 100% with both low and high cutting depths. The results indicated the robust and reliable wheel wear detection in cylindrical grinding with the use of relatively cheap sensors like accelerometers. The proposed method can be widely used in many applications in the industry where grinding is predominantly used as the finishing operation.