No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Biofouling Detection and Extent Classification in Tidal Stream Turbines via a Soft Voting Ensemble Transfer Learning Approach
Abstract
This paper investigates the use of transfer learning models for detecting and classifying biofouling extent in tidal stream turbines. It provides a comprehensive exploration of a soft voting ensemble approach to deal with a multiclass image detection problem. In this context, three well-known convolutional neural network models, namely VGG16, ResNet50, and MobileNetV2, are the primary focus of the proposed study and are trained on an experimental dataset issued from the Shanghai Maritime University tidal stream turbine platform.
... Over time, a complex threedimensional community is created on the substrate, as shown in Figure. 1 [1]. Turbine Technology Impact on Biofouling Accumulation: Biofouling affects tidal stream turbine performance by increasing drag and causing recirculation loops and vertices, even with partial colonization [2]. It can also damage the rotor and accelerate corrosion of the protective layer on the blade ( Figure. ...
... Illustration of the corrosion acceleration due to biofouling. © Courtesy of Prof. Yusaku Kyozuka[2]. ...
... Illustration of highly impacting biofouling. Courtesy of Prof. Yusaku Kyozuka[2] ...
This study uses a transfer learning-based methodology and a soft voting ensemble approach with CNNs (VGG, ResNet, MobileNet) to address biofouling in tidal stream turbines. Experimental results show 90% accuracy, enhancing detection and classification for optimized turbine operation.
... These algorithms can learn to extract relevant features from processed images and classify biofouling instances with greater accuracy. By training models on labeled datasets of biofouling images, machine learning algorithms can improve their ability to detect and classify biofouling, even in conditions of limited visibility [12]. ...
... The paper [12] presents a novel method for detecting and classifying biofouling on TSTs using a soft voting ensemble transfer learning approach. Through the utilization of pretrained deep learning models and data augmentation techniques, the proposed method effectively identifies and categorizes biofouling levels on turbine surfaces. ...
Biofouling poses a significant challenge to the efficiency and performance of tidal stream turbines. This study proposes a novel approach utilizing Fast R-CNN for biofouling detection and coverage estimation in tidal stream turbines. We use a real biofouling dataset collected from four turbines installed on the pier of Ikitsuki Bridge in Hirado, Nagasaki, Japan, over 8 months, containing both clean and biofouled turbine images. Employing various augmentation techniques, we enhanced the dataset's robustness. Our method achieved a remarkable 94% accuracy in detecting and classifying biofouled turbine blades. Additionally, using the OpenCV library in Python, we estimated the extent of biofouling coverage on turbine blades, offering valuable insights for maintenance and optimization efforts and assisting informed decision-making on strategies for tidal stream turbine maintenance and performance optimization.
... Over time, a complex threedimensional community is created on the substrate, as shown in Figure. 1 [1]. Turbine Technology Impact on Biofouling Accumulation: Biofouling affects tidal stream turbine performance by increasing drag and causing recirculation loops and vertices, even with partial colonization [2]. It can also damage the rotor and accelerate corrosion of the protective layer on the blade ( Figure. ...
... Illustration of the corrosion acceleration due to biofouling. © Courtesy of Prof. Yusaku Kyozuka[2]. ...
This research proposes to map a machine learning path forward for tidal stream turbines biofouling detection and estimation. The proposed review covers an overview of biofoul-ing and its impact on tidal stream turbines, current techniques for detecting and estimating biofouling, recent developments , and challenges in the field, as well as several promising prospects for biofouling detection and estimation..
... Recent frameworks have exploited image-based methods, which are the easiest to master and the most efficient, despite the disadvantage of noisy and blurry images (Rashid et al., 2023b). Bloomfield et al. (2021) applied deep learning to automate the classification of biofouling in images and compared it to expert assessments. ...
... This procedure involves using networks trained on complex classification tasks and adapting them to achieve a less complex task (Anaya-Isaza and Mera-Jiménez, 2022). Several convolution networks are used for this purpose, such as ResNet50, VGG16, VGG19, etc. (Rashid et al., 2023b). These networks were trained on a large dataset developed for standard computer vision benchmarks. ...
Harnessing the power of tidal streams is a sustainable way of exploiting renewable marine energy resources. It involves installing tidal stream turbines underwater to harness the energy. Nevertheless, these turbines are prone to the accumulation of biofouling, which significantly reduces their energy output and operational efficiency. It is therefore crucial to implement a condition-based monitoring system to detect biofouling promptly and ensure the continuous operation of a tidal stream turbine. In this context, this paper presents a data-centric approach that uses model submerged tidal stream turbine video images to detect and quantify biofouling. The relevance of a two-dimensional variational mode decomposition approach is investigated to extract relevant information from the potentially noisy collected images. While generative adversarial networks are used to address the data imbalance problem, a convolutional neural network is adopted to detect and assess the extent of biofouling. The performance of the proposed approach is assessed and validated using two experimental datasets obtained from the tidal stream turbine platforms of the Shanghai Maritime University and the Lehigh University.
... However, in tidal turbines, it's especially problematic due to the harsh marine conditions. Constant exposure to seawater and swift tidal currents creates an ideal environment for marine growth [1]. ...
... Various methods have been developed to detect biofouling on TSTs (Chin et al., 2017;Rashid et al., 2023a;Zheng et al., 2019). For instance, sparse autoencoding combined with softmax regression has shown promise for real-time monitoring (Zheng et al., 2019), while depthwise separable convolutional neural networks (CNNs) have been utilized to reduce computational complexity while capturing local spatial relationships (Xin et al., 2021). ...
This study addresses the biofouling challenges in Tidal Stream Turbines (TSTs) to ensure their reliable and optimal operation. In this context, it is proposed an effective methodology employing a soft voting ensemble transfer learning-based approach for the detection and extent classification of biofouling. The proposed framework incorporates essential components such as data augmentation and pre-processing, including image resizing and data segmentation, forming a comprehensive video image-based approach. To overcome the constraint of limited data, experimental investigations were conducted, resulting in the acquisition of two datasets: one from the TST platform at Shanghai Maritime University (SMU) and the other from the tidal turbulence test facility at Lehigh University (LU). The three prominent convolutional neural network models, namely Visual Geometry Group (VGG), Residual Network (ResNet) and MobileNet, trained on these datasets, demonstrate precise detection and classification of turbine conditions, achieving an accuracy of 83% for the SMU dataset and 90% for the LU dataset. The noted disparity in accuracy for the SMU dataset is attributed to its smaller size, highlighting the significant impact of dataset scale on classification performance. This study provides valuable insight into the development of effective biofouling detection and classification strategies for TST systems.
... Given the brief overview of the current state of the art and the roadmap outlined in [18], it becomes evident that monitoring biofouling in TSTs through image processing demands an appropriate machine learning approach [19]. This approach must effectively tackle challenges such as limited data availability [20,21], enhance generalization and robustness, and expedite the learning process. ...
In the context of harvesting tidal stream energy, which is considered a promising source of renewable energy due to its high energy density, stability, and predictability, this paper proposes a review-based roadmap investigating the use of data-driven techniques, more specifically machine learning-based approaches, to detect and estimate the extent of biofouling in tidal stream turbines. An overview of biofouling and its impact on these turbines will be provided as well as a brief review of current methodologies and techniques for detecting and estimating biofouling. Additionally, recent developments and challenges in the field will be examined, while providing several promising prospects for biofouling detection and estimation in tidal stream turbines.
Marine current turbines (MCTs) may exhibit reduced energy production and structural instability due to attachments, such as biofouling and plankton. Semantic segmentation (SS) is utilized to recognize these attachments, enabling on-demand maintenance towards optimizing power generation efficiency and minimizing maintenance costs. However, the degree of motion blur might vary according to the MCT rotational speed. The SS methods are not robust against such variations, and the recognition accuracy could be significantly reduced. In order to alleviate this problem, the SS method is proposed based on image entropy weighted spatio-temporal fusion (IEWSTF). The method has two features: (1) A spatio-temporal fusion (STF) mechanism is proposed to learn spatio-temporal (ST) features in adjacent frames while conducting feature fusion, thus reducing the impact of motion blur on feature extraction. (2) An image entropy weighting (IEW) mechanism is proposed to adjust the fusion weights adaptively for better fusion effects. The experimental results demonstrate that the proposed method achieves superior recognition performance with MCT datasets with various rotational speeds and is more robust to rotational speed variations than other methods.
Accurate bearing fault diagnosis is of great significance of the safety and reliability of rotary mechanical system. In practice, the sample proportion between faulty data and healthy data in rotating mechanical system is imbalanced. Furthermore, there are commonalities between the bearing fault detection, classification, and identification tasks. Based on these observations, this article proposes a novel integrated multitasking intelligent bearing fault diagnosis scheme with the aid of representation learning under imbalanced sample condition, which realizes bearing fault detection, classification, and unknown fault identification. Specifically, in the unsupervised condition, a bearing fault detection approach based on modified denoising autoencoder (DAE) with self-attention mechanism for bottleneck layer (MDAE-SAMB) is proposed in the integrated scheme, which only uses the healthy data for training. The self-attention mechanism is introduced into the neurons in the bottleneck layer, which can assign different weights to the neurons in the bottleneck layer. Moreover, the transfer learning based on representation learning is proposed for few-shot fault classification. Only a few fault samples are used for offline training, and high-accuracy online bearing fault classification is achieved. Finally, according to the known fault data, the unknown bearing faults can be effectively identified. A bearing dataset generated by rotor dynamics experiment rig (RDER) and a public bearing dataset demonstrates the applicability of the proposed integrated fault diagnosis scheme.
In this study, we propose a method for inspecting the condition of hull surfaces using underwater images acquired from the camera of a remotely controlled underwater vehicle (ROUV). To this end, a soft voting ensemble classifier comprising six well-known convolutional neural network models was used. Using the transfer learning technique, the images of the hull surfaces were used to retrain the six models. The proposed method exhibited an accuracy of 98.13%, a precision of 98.73%, a recall of 97.50%, and an F1-score of 98.11% for the classification of the test set. Furthermore, the time taken for the classification of one image was verified to be approximately 56.25 ms, which is applicable to ROUVs that require real-time inspection.
The power coefficient for a horizontal axis tidal turbine is the determinant factor for the efficiency of a tidal energy system. To guarantee a highly efficient tidal turbine operating in the real sea environment for an enduring long period is of critical importance to the power production and hence the cost of energy. However, this performance is under the threat of marine biofouling and the biofouling effect on tidal turbine systems are barely known neither quantified. This paper focuses on the study of the roughness effect due to biofouling on the performance of a tidal turbine. A Reynolds Averaged Navier-Stokes model based Computational Fluid Dynamics (CFD) was developed to predict the effect of biofouling on a full-scale turbine. A roughness modelling that involves modified wall-functions in the CFD model was used representing the surface roughness caused by barnacle fouling. The simulations were conducted under different fouling scenarios for a range of tip speed ratios (TSR). The surface fouling resulted in up to 13% decrease in the power coefficient at the designed operating condition. The effect proved to be even more severe at higher TSRs, bringing narrower operating range of TSRs. The results also suggest that by lowering the operating TSRs for fouled turbines the fouling effect on efficiency losses can be minimised to ensure efficient operation between maintenances.
Most electrical machines and drive signals are non-Gaussian and are highly nonlinear in nature. A useful set of techniques to examine such signals relies on higher-order statistics (HOS) spectral representations. They describe statistical dependencies of frequency components that are neglected by traditional spectral measures, namely the power spectrum (PS). One of the most used HOS is the bispectrum where examining higher-order correlations should provide further details and information about the conditions of electric machines and drives. In this context, the stator currents of electric machines are of particular interest because they are periodic, nonlinear, and cyclostationary. This current is, therefore, well adapted for analysis using bispectrum in the designing of an efficient condition monitoring method for electric machines and drives. This paper is, therefore, proposing a bispectrum-based diagnosis method dealing the with tidal stream turbine (TST) rotor blades biofouling issue, which is a marine environment natural process responsible for turbine rotor unbalance. The proposed bispectrum-based diagnosis method is verified using experimental data provided from a permanent magnet synchronous generator (PMSG)-based TST experiencing biofouling emulated by attachment on the turbine blade. Based on the achieved results, it can be concluded that the proposed diagnosis method has been very successful. Indeed, biofouling imbalance-related frequencies are clearly identified despite marine environmental nuisances (turbulences and waves).
Most electrical machines and drive signals are non-Gaussian and are highly nonlinear in nature. A useful set of techniques to examine such signals relies on higher-order statistics (HOS) spectral representations. They describe statistical dependencies of frequency components that are neglected by traditional spectral measures, namely the power spectrum (PS). One of the most used HOS is the bispectrum where examining higher-order correlations should provide further details and information about the conditions of electric machines and drives. In this context, the stator currents of electric machines are of particular interest because they are periodic, nonlinear, and cyclostationary. This current is, therefore, well adapted for analysis using bispectrum in the designing of an efficient condition monitoring method for electric machines and drives. This paper is, therefore, proposing a bispectrum-based diagnosis method dealing the with tidal stream turbine (TST) rotor blades biofouling issue, which is a marine environment natural process responsible for turbine rotor unbalance. The proposed bispectrum-based diagnosis method is verified using experimental data provided from a permanent magnet synchronous generator (PMSG)-based TST experiencing biofouling emulated by attachment on the turbine blade. Based on the achieved results, it can be concluded that the proposed diagnosis method has been very successful. Indeed, biofouling imbalance-related frequencies are clearly identified despite marine environmental nuisances (turbulences and waves).
The development and application of marine current energy are attracting more and more attention around the world. Due to the hardness of its working environment, it is important and difficult to study the fault diagnosis of a marine current generation system. In this paper, an underwater image is chosen as the fault-diagnosing signal, after different sensors are compared. This paper proposes a diagnosis method based on the sparse autoencoder (SA) and softmax regression (SR). The SA is used to extract the features and SR is used to classify them. Images are used to monitor whether the blade is attached by benthos and to determine its corresponding degree of attachment. Compared with other methods, the experiment results show that the proposed method can diagnose the blade attachment with higher accuracy.
The control of biofouling on marine vessels is challenging and costly. Early detection before hull performance is significantly affected is desirable, especially if “grooming” is an option. Here, a system is described to detect marine fouling at an early stage of development. In this study, an image of fouling can be transferred wirelessly via a mobile network for analysis. The proposed system utilizes transfer learning and deep convolutional neural network (CNN) to perform image recognition on the fouling image by classifying the detected fouling species and the density of fouling on the surface. Transfer learning using Google’s Inception V3 model with Softmax at last layer was carried out on a fouling database of 10 categories and 1825 images. Experimental results gave acceptable accuracies for fouling detection and recognition.
Most supervised learning-based approaches follow the assumptions that offline data and online data must obey a similar distribution, which is difficult to satisfy in realistic remaining useful life (RUL) prediction. To solve the problem, domain adaptation (DA) learning-oriented transfer learning (TL) was proposed. Nevertheless, only adopting a conventional global DA approach may confuse the fine-grained features between subdomains represented by different degenerate stages. Consequently, a novel variational auto-encoder-long short-term memory network-local weighted deep sub-domain adaptation network (VLSTM-LWSAN) is proposed for RUL prediction. Specifically, the input data are compressed into the interpretable latent space, from which the fine-grained features between subdomains are local alignment through local weighted deep sub-domain adaptation network. In this sense, the discrepancy between the unlabeled target domain and the source domain is decreased. The proposed VLSTM-LWSAN is verified by an aircraft turbofan engine dataset. The research results represent that the VLSTM-LWSAN outperforms some deep learning approaches without transfer learning and conventional transfer learning approaches.
Although several problems related to biofouling of marine current turbines (MCTs) are reported in the literature, the most important one is related to long-term operational performance and maintenance costs. Nevertheless, studies related to the impact of biofouling on MCT performance are rather scarce. In this study, the impact of biofilm on MCT performance is investigated using the Computational Fluid Dynamics (CFD) approach. Biofilm is modelled using previously developed roughness functions implemented in a wall function solver. A verification study is performed to determine sufficient grid spacings and to calculate numerical uncertainty. The validation study is conducted by comparing the obtained results with experimental and numerical ones from the literature. Full-scale CFD simulations are performed for six fouling conditions with varying biofilm height and percentage of surface coverage at eight tip speed ratios (TSRs). The obtained results highlight the significant impact of biofilm on MCT performance reflected in a decrease in the power coefficient, which for the optimal TSR is equal to −10.7% for the R1 fouling condition. Finally, a detailed analysis of the flow around MCTs fouled with biofilm is conducted and the reasons for the detrimental impact of biofilm on MCT performance are discussed.
Large-scale pre-trained models (PTMs) such as BERT and GPT have recently achieved great success and become a milestone in the field of artificial intelligence (AI). Owing to sophisticated pre-training objectives and huge model parameters, large-scale PTMs can effectively capture knowledge from massive labeled and unlabeled data. By storing knowledge into huge parameters and fine-tuning on specific tasks, the rich knowledge implicitly encoded in huge parameters can benefit a variety of downstream tasks, which has been extensively demonstrated via experimental verification and empirical analysis. It is now the consensus of the AI community to adopt PTMs as backbone for downstream tasks rather than learning models from scratch. In this paper, we take a deep look into the history of pre-training, especially its special relation with transfer learning and self-supervised learning, to reveal the crucial position of PTMs in the AI development spectrum. Further, we comprehensively review the latest breakthroughs of PTMs. These breakthroughs are driven by the surge of computational power and the increasing availability of data, towards four important directions: designing effective architectures, utilizing rich contexts, improving computational efficiency, and conducting interpretation and theoretical analysis. Finally, we discuss a series of open problems and research directions of PTMs, and hope our view can inspire and advance the future study of PTMs.
To diagnose the attachment of marine current turbine, this article proposes a method based on convolutional neural network and the concepts of depthwise separable convolution to achieve feature extraction. The method consists of three steps: data preprocessing, feature extraction and fault diagnosis. This method can diagnose the fault degree of blade imbalance and uniform attachment in underwater environment with strong currents and complex spatiotemporal variability. It can extract distinct image feature in harsh marine environments by using a convolutional neural network. In addition, this method is robust for the recognition of blurred pictures under high-speed rotation.
Deep neural networks are powerful, but using these networks is both memory and time consuming due to their numerous parameters and large amounts of computation. Many studies have been conducted on compressing the models on the parameter-level as well as on the bit-level. Here, we propose an efficient strategy to compress on the layers that are computation or memory consuming. We compress the model by introducing global average pooling, performing iterative pruning on the filters with the proposed order-deciding scheme in order to prune more efficiently, applying truncated SVD to the fully-connected layer, and performing quantization. Experiments on the VGG16 model show that our approach achieves a 60.9 × compression ratio in off-line storage with about 0.848% and 0.1378% loss of accuracy in the top-1 and top-5 classification results, respectively, with the validation dataset of ILSVRC2012. Our approach also shows good compression results on AlexNet and faster R-CNN.
In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively.
In recent years, several studies have been conducted to evaluate damages caused by biofouling that results in organisms attachment on a surface in contact with water for a period of time. As presented, the biofouling problem seems quite simple. Unfortunately, there are a multitude of organisms that cause biofouling as well as several types of affected surfaces and, consequently, lead to different solutions to address the biofouling problem. In a marine renewable energies context, the biofouling issues facing submerged systems are subject to the same settling processes. This paper is therefore an attempt to provide an exhaustive state of the art review, which allow assessing the biofouling problem in marine renewable energy converters (MREC). The proposed review will specifically highlight the marine environment impacts and the solutions to prevent fouling. Challenges faced by the marine renewable energies market for massive deployment are also discussed.
Deeper neural networks are more difficult to train. We present a residual
learning framework to ease the training of networks that are substantially
deeper than those used previously. We explicitly reformulate the layers as
learning residual functions with reference to the layer inputs, instead of
learning unreferenced functions. We provide comprehensive empirical evidence
showing that these residual networks are easier to optimize, and can gain
accuracy from considerably increased depth. On the ImageNet dataset we evaluate
residual nets with a depth of up to 152 layers---8x deeper than VGG nets but
still having lower complexity. An ensemble of these residual nets achieves
3.57% error on the ImageNet test set. This result won the 1st place on the
ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100
and 1000 layers.
The depth of representations is of central importance for many visual
recognition tasks. Solely due to our extremely deep representations, we obtain
a 28% relative improvement on the COCO object detection dataset. Deep residual
nets are foundations of our submissions to ILSVRC & COCO 2015 competitions,
where we also won the 1st places on the tasks of ImageNet detection, ImageNet
localization, COCO detection, and COCO segmentation.
In this work we investigate the effect of the convolutional network depth on
its accuracy in the large-scale image recognition setting. Our main
contribution is a thorough evaluation of networks of increasing depth, which
shows that a significant improvement on the prior-art configurations can be
achieved by pushing the depth to 16-19 weight layers. These findings were the
basis of our ImageNet Challenge 2014 submission, where our team secured the
first and the second places in the localisation and classification tracks
respectively.
Global renewable energy production
- I E Agency
I. E. Agency, "Global renewable energy production," Accessed on April
25 th, 2023, data source: International Energy Agency.