Search results

Faults in the actual outdoor performance of Building Integrated Photovoltaic (BIPV) systems can go unnoticed for several months since the energy productions are subject to significant variations that could mask faulty behaviors. Even large BIPV energy deficits could be hard to detect. The purpose of this paper is to develop a cost-effective approach to automatically detect faults in the energy productions of BIPV systems using historical BIPV energy productions as the only source of information that is typically collected in all BIPV systems.

Energy productions of BIPV systems are time series in nature. Therefore, time series methods are used to automatically detect two categories of faults (outliers and structure changes) in the monthly energy productions of BIPV systems. The research methodology consists of the automatic detection of outliers in energy productions, and automatic detection of structure changes in energy productions.

The proposed approach is applied to detect faults in the monthly energy productions of 89 BIPV systems. The results confirm that outliers and structure changes can be automatically detected in the monthly energy productions of BIPV systems using time series methods in presence of short-term variations, monthly seasonality, and long-term degradation in performance.

Unlike existing methods, the proposed approach does not require performance ratio calculation, operating condition data, such as solar irradiation, or the output of neighboring BIPV systems. It only uses the historical information about the BIPV energy productions to distinguish between faults and other time series properties including seasonality, short-term variations, and degradation trends.

Reports how the theorem of the texture “tuned” mask was modified to solve some problems encountered in the automatic faults (including filling bars, oil stains, weft‐lacking and holes) detection and recognition of the plain woven fabrics. These problems are the faults of variable shapes and sizes, those of variable structure and the grey‐level differences in the faults of oil stains. The index of the “tuned” mask in the texture “tuned” mask theorem was modified to converge the variability of the faults, and to elongate the distances between each fault’s average texture energy so that the texture energy in normal texture and in faults can be confined to different fixed ranges. The results show that the optimum texture “tuned” mask found from the modified theorem of the texture “tuned” mask can be used satisfactorily to identify different faults due to structure, shapes and size variation. However, in the case of undertoned oil stains and lower density filling bars, this method may sometimes cause misidentification.

This study aims to bring awareness to the developing of fault detection systems using the data collected from sensor devices/physical devices of various systems for predictive maintenance. Opportunities and challenges in developing anomaly detection algorithms for predictive maintenance and unexplored areas in this context are also discussed.

For conducting a systematic review on the state-of-the-art algorithms in fault detection for predictive maintenance, review papers from the years 2017–2021 available in the Scopus database were selected. A total of 93 papers were chosen. They are classified under electrical and electronics, civil and constructions, automobile, production and mechanical. In addition to this, the paper provides a detailed discussion of various fault-detection algorithms that can be categorised under supervised, semi-supervised, unsupervised learning and traditional statistical method along with an analysis of various forms of anomalies prevalent across different sectors of industry.

Based on the literature reviewed, seven propositions with a focus on the following areas are presented: need for a uniform framework while scaling the number of sensors; the need for identification of erroneous parameters; why there is a need for new algorithms based on unsupervised and semi-supervised learning; the importance of ensemble learning and data fusion algorithms; the necessity of automatic fault diagnostic systems; concerns about multiple fault detection; and cost-effective fault detection. These propositions shed light on the unsolved issues of predictive maintenance using fault detection algorithms. A novel architecture based on the methodologies and propositions gives more clarity for the reader to further explore in this area.

Papers for this study were selected from the Scopus database for predictive maintenance in the field of fault detection. Review papers published in this area deal only with methods used to detect anomalies, whereas this paper attempts to establish a link between different industrial domains and the methods used in each industry that uses fault detection for predictive maintenance.

The purpose of this paper is to test the applicability and the performance of an approach called logical analysis of data (LAD) on the detection of faults in rotating machinery using vibration signals.

LAD is a supervised learning data mining technique that relies on finding patterns in a binary database to generate decision functions. The hypothesis is that a LAD‐based decision model can be used as an effective tool for automatic detection of faults in rolling element bearings. A novel Multiple Integer Linear Programming approach is used to generate patterns for the LAD decision model. Frequency and time‐based features are extracted from rotor bearing vibration signals and are pre‐processed to be suitable for use with LAD.

The results show good classification accuracy with both time and frequency features.

The diagnostic tool implemented in the form of software in a production or operations maintenance environment can be very helpful to maintenance experts as it reveals the patterns that lead to the diagnosis in interpretable terms which facilitates efforts to understand the reasons behind the components' failure.

The proposed modifications to the LAD‐based decision model which is being tested for the first time in the field of fault detection in rotating machinery lead to improved accuracy results in addition to the added value of result interpretability due to this distinctive property of LAD.

This paper aims to improve the performance and speed of artificial neural network (ANN)‐ball‐bearing fault detection expert systems by eliminating unimportant inputs and changing the ANN structure.

An algorithm is used to select the best subset of features to boost the success of detecting healthy and faulty ball. Some of the important parameters of the ANN are also optimized to make the classifier achieve the maximum performance.

It was found that better accuracy can be obtained for ANN with fewer inputs.

The method can be used for other machinery condition‐monitoring systems which are based on ANN.

The results are useful for bearing fault detection systems designers and quality check centers in bearing manufacturing companies.

The algorithm used in this research is faster than in previous studies. Changing ANN parameters improved the results. The system was examined using experimental data of ball‐bearings.

This paper aims to investigate methods of implementing in‐process fault avoidance in robotic friction stir welding (FSW).

Investigations into the possibilities for automatically detecting gap‐faults in a friction stir lap weld were conducted. Force signals were collected from a number of lap welds containing differing degrees of gap faults. Statistical analysis was carried out to determine whether these signals could be used to develop an automatic fault detector/classifier.

The results demonstrate that the frequency spectra of collected force signals can be mapped to a lower dimension through discovered discriminant functions where the faulty welds and control welds are linearly separable. This implies that a robust and precise classifier is very plausible, given force signals.

Future research should focus on a complete controller using the information reported in this paper. This should allow for a robotic friction stir welder to detect and avoid faults in real time. This would improve manufacturing safety and yield.

This paper is applicable to the rapidly expanding robotic FSW industry. A great advantage of heavy machine tool versus robotic FSW is that the robot cannot supply the same amount of rigidity. Future work must strive to overcome this lack of mechanical rigidity with intelligent control, as has been examined in this paper.

This paper investigates fault detection in robotic FSW. Fault detection and avoidance are essential for the increased robustness of robotic FSW. The paper's results describe very promising directions for such implementation.

This paper aims to provide a novel integrated fault detection method for industrial process monitoring.

A novel integrated fault detection method based on the combination of Mallat (MA) algorithm, weight-elimination (WE) algorithm, conjugate gradient (CG) algorithm and multi-dimensional Taylor network (MTN) dynamic model, namely, MA-WE-CG-MTN, is proposed in this paper. First, MA algorithm is taken as data pre-processing. Second, in virtue of approximation ability and low computation complexity owing to the simple structure of MTN, MTN dynamic models are constructed for each frequency band. Furthermore, the CG algorithm is used to discipline the model parameters and the outputs of MTN model of each frequency band are gained. Third, the authors introduce the WE algorithm to cut down the number of middle layer nodes of MTN, reducing the complexity of the network. Finally, the outputs of MTN model for each frequency band are superimposed to achieve outputs of MTN model, and fault detection is proceeded by the residual error generator based on the difference between the output of MTN model and the actual output.

The novel proposed method is used to perform fault detection for industrial process monitoring effectively, such as the Benchmark Simulation Model 1 wastewater treatment process.

The novel proposed method has generality and provides considerably improved performance and effectiveness, which is used to perform fault detection for industrial process monitoring. The proposed method has good robustness, low complexity and easy implementation.

The purpose of this paper is to study the fault diagnosis of internal combustion (IC) engine gearbox using vibration signals with signal processing and machine learning (ML) techniques.

Vibration signals from the gearbox are acquired for healthy and induced faulty conditions of the gear. In this study, 50% tooth fault and 100% tooth fault are chosen as gear faults in the driver gear. The acquired signals are processed and analyzed using signal processing and ML techniques.

The obtained results show that variation in the amplitude of the crankshaft rotational frequency (CRF) and gear mesh frequency (GMF) for different conditions of the gearbox with various load conditions. ML techniques were also employed in developing the fault diagnosis system using statistical features. J48 decision tree provides better classification accuracy about 85.1852% in identifying gearbox conditions.

The proposed approach can be used effectively for fault diagnosis of IC engine gearbox. Spectrum and continuous wavelet transform (CWT) provide better information about gear fault conditions using time–frequency characteristics.

In this paper, experiments are conducted on real-time running condition of IC engine gearbox while considering combustion. Eddy current dynamometer is attached to output shaft of the engine for applying load. Spectrum, cepstrum, short-time Fourier transform (STFT) and wavelet analysis are performed. Spectrum, cepstrum and CWT provide better information about gear fault conditions using time–frequency characteristics. ML techniques were used in analyzing classification accuracy of the experimental data to detect the gearbox conditions using various classifiers. Hence, these techniques can be used for detection of faults in the IC engine gearbox and other reciprocating/rotating machineries.

The brake controller is a key component of the locomotive brake system. It is essential to study its safety.

This paper summarizes and analyzes typical faults of the brake controller, and proposes four categories of faults: position sensor faults, microswitch faults, mechanical faults and communication faults. Suggestions and methods for improving the safety of the brake controller are also presented.

In this paper, a self-judgment and self-learning dynamic calibration method is proposed, which integrates the linear error of the sensor and the manufacturing and assembly errors of the brake controller to solve the output drift. This paper also proposes a logic for diagnosing and handling microswitch faults. Suggestions are proposed for other faults of brake controller.

The methods proposed in this paper can greatly improve the usability of the brake controller and reduce the failure rate.

This paper aims to design an AI-based drone that can facilitate the complicated and time-intensive control process for detecting healthy and defective solar panels. Today, the use of solar panels is becoming widespread, and control problems are increasing. Physical control of the solar panels is critical in obtaining electrical power. Controlling solar panel power plants and rooftop panel applications installed in large areas can be difficult and time-consuming. Therefore, this paper designs a system that aims to panel detection.

This paper designed a low-cost AI-based unmanned aerial vehicle to reduce the difficulty of the control process. Convolutional neural network based AI models were developed to classify solar panels as damaged, dusty and normal. Two approaches to the solar panel detection model were adopted: Approach 1 and Approach 2.

The training was conducted with YOLOv5, YOLOv6 and YOLOv8 models in Approach 1. The best F1 score was 81% at 150 epochs with YOLOv5m. In total, 87% and 89% of the best F1 score and mAP values were obtained with the YOLOv5s model at 100 epochs in Approach 2 as a proposed method. The best models at Approaches 1 and 2 were used with a developed AI-based drone in the real-time test application.

The AI-based low-cost solar panel detection drone was developed with an original data set of 1,100 images. A detailed comparative analysis of YOLOv5, YOLOv6 and YOLOv8 models regarding performance metrics was realized. Gaussian, salt-pepper noise addition and wavelet transform noise removal preprocessing techniques were applied to the created data set under the proposed method. The proposed method demonstrated expressive and remarkable performance in panel detection applications.