Enhancing Weather Monitoring for Agriculture with Deep Learning: Anomaly Detection in East Java Using LSTM Autoencoder and OCSVM

By pushing computing resources from the cloud to the network edge close to mobile users, mobile edge computing (MEC) enables low latency for a wide variety of applications. Nevertheless, in dynamic MEC systems, MEC services are challenged by the risks of runtime reliability anomalies. Detecting runtime reliability anomalies for MEC services is challenging yet critical to ensuring the stability of MEC systems. The effectiveness of existing anomaly detection methods suffers from poor performance when handling MEC services' large-volume, continuous, and volatile reliability streaming data. The key is to identify significant changes in the distribution of MEC services' current reliability streaming data compared with their historical performance. Inspired by concept drift, this paper proposes B-Detection, a boosting Long Short-Term Memory (LSTM) Autoencoder for detecting MEC services' runtime reliability anomalies based on distribution dissimilarity evaluation. B-Detection employs a deep learning method named LSTM Autoencoder to characterize the MEC services' historical reliability data distribution. To cope with the challenge of modeling complex distribution characteristics of MEC services' historical reliability streaming data and guarantee the real-time performance of B-Detection, we enhance LSTM Autoencoder with a weight-based reservoir sampling technique and an LSTM boosting algorithm. The reconstruction loss of the trained LSTM Autoencoder model is estimated for the up-to-date reliability streaming data, and the result is used to infer MEC services' runtime reliability anomalies. The performance of B-Detection is verified through a series of experiments conducted on a real-world dataset.

Early fault diagnosis of equipment based on the current condition assessment is one of the commonly used methods of CBM (condition-based maintenance). It refers to mining the impending fault characteristics from a large amount of production data in long-term operation (Luo et al., 2019). However, these data are huge multivariate data causing a difficulty in extracting features manually. In manufacturing scenario, the majority of machines are in a normal state and the abnormalities are relatively rare which makes the collected data occurring data imbalancing.This study explores the use of LSTM (Long Short-Term Memory) autoencoder combined with DBSCAN (Density-based spatial clustering of applications with noise) under the condition of data imbalance. The reconstruction error of the model after training is used as an evaluation index where the errors of each time point between the reconstruction sequence and the actual sequence are calculated and inputted for classification in the DBSCAN model.In this study, a water distribution system dataset from the SKAB (Skoltech Anomaly Benchmark) was used to verify the anomaly detection of our proposed model. Our model shows the F1-score of 0.8025 which is better than the four models proposed by Moon et al. in 2023. With a LSTM autoencoder, the proposed DBSCAN classification model can avoid the difficulty of setting a threshold value in classification.

The increasing demand for sustainable aquaculture solutions has accelerated interest in microalgae as an alternative fish feed source. This study presents a novel integration of Long-Range Wide Area Network (LoRaWAN) sensors and a Long Short-Term Memory (LSTM) autoencoder model to enable real-time monitoring and anomaly detection in 300L outdoor microalgae cultivation. A solar-powered Internet of Things (IoT) sensor and LoRaWAN system were developed to monitor key water quality parameters such as pH, water temperature, electrical conductivity (EC), and oxidation–reduction potential (ORP). ORP was identified as a sensitive alternative for monitoring microalgae growth dynamics and contamination. The LSTM autoencoder, trained on normal ORP data, effectively distinguished abnormal culture patterns based on reconstruction errors, achieving high precision (1.000), recall (0.944), F1-score (0.971), and AUC-ROC (1.000). The addition of a moving average filter improved data stability for model training of filtered sensor signals. A bespoke visual reconstruction error heatmap further simplified anomaly interpretation for end-users. This framework advances smart aquaculture by enabling timely intervention, supporting sustainable microalgae production, and aligning with global food security and environmental sustainability goals.

Anomaly detection is critical in various sectors, offering significant advantages by precisely identifying and mitigating system failures and errors, thus preventing severe losses. This study provides a comprehensive comparative anomaly detection analysis through two sophisticated deep learning models: Autoencoder and Long Short-Term Memory (LSTM) Autoencoder, explicitly focusing on temperature and sound time series data. The paper starts with a detailed theoretical foundation, elaborating on both models' mechanics and mathematical formulations. We then advance to the empirical phase, where these models are rigorously trained and tested against a robust dataset. The effectiveness of each model is meticulously assessed through a suite of metrics that gauge their accuracy, sensitivity, and robustness in anomaly detection scenarios. Additionally, we explore the deployment of these models in a real-time environment, where they actively engage in anomaly detection on incoming data streams. The anomalies detected are dynamically displayed on a user-friendly graphical interface, making the results readily accessible and interpretable for users at all levels of technical expertise. Quantitative evaluations of the models are conducted using key performance metrics such as accuracy, precision, recall, and F1-score. Our analysis reveals that the LSTM Autoencoder model excels with an impressive accuracy rate of 99%, while other metrics also affirm its superior performance, marking it as exceptionally effective and reliable. This study highlights the LSTM Autoencoder's advanced anomaly detection capabilities and establishes its superiority over the traditional Autoencoder model in processing complex time series data. The insights gained here are crucial for industries focused on predictive maintenance and quality control, where early anomaly detection is key to maintaining operational efficiency and safety.

The major challenge in predictive maintenance (PdM) is an insufficiency of failure data used to train the model and the high complexity of industrial plants, where operational conditions changed over time hence the predetermined threshold will change frequently. Novelty detection offers a solution to this problem by detecting anomalies through learning only the normal data. This study implements a novelty detection method as a one-class classification to detect early signs of failure in a rolling bearing using long short-term memory (LSTM) autoencoder. LSTM autoencoder is a deep learning algorithm combining an autoencoder and LSTM network to reconstruct the normal data to learn nonlinear relationships and temporal nature. The model was tested on historical multivariate time-series fault data provided by Case Western Reserve University. The dropout layer is implemented to reduce overfitting by creating an ensemble model of a neural network. The results suggested that the LSTM autoencoder can effectively differentiate between normal and fault patterns of the bearing with up to 92 % accuracy.

The problem of energy depletion has brought wind energy under consideration to replace oil- or chemical-based energy. However, the breakdown of wind turbines is a major concern. Accordingly, unsupervised learning was performed using the vibration signal of a wind power generator to achieve an outlier detection performance of 97%. We analyzed the vibration data through wavelet packet conversion and identified a specific frequency band that showed a large difference between the normal and abnormal data. To emphasize these specific frequency bands, high-pass filters were applied to maximize the difference. Subsequently, the dimensions of the data were reduced through principal component analysis, giving unique characteristics to the data preprocessing process. Normal data collected from a wind farm located in northern Sweden was first preprocessed and trained using a long short-term memory (LSTM) autoencoder to perform outlier detection. The LSTM Autoencoder is a model specialized for time-series data that learns the patterns of normal data and detects other data as outliers. Therefore, we propose a method for outlier detection through data preprocessing and unsupervised learning, utilizing the vibration signals from wind generators. This will facilitate the quick and accurate detection of wind power generator failures and provide alternatives to the problem of energy depletion.

Land monitoring is important in agriculture. Early warning information regarding the land condition enable farmers to respond quickly when anomaly condition occures. However, identifying anomaly of land condition is not a simple task. In this research, a model of anomaly detection for land monitoring system is proposed. Raw data collected from land monitoring sensors is used as the dataset. Isolation Forest is used to transform the unlabeled data into labeled data. The labeled dataset is then used to create anomaly detection model using Long Short-Term Memory (LSTM) autoencoder. The experiments results show that the Isolation Forest has the potential to label data. The LSTM autoencoder has the accuracy 0.95 precision 0.96, recall 0.99 and flscore 0.97.

Noise reduction is one of the most important process used for signal processing in communication systems. The signal-to-noise ratio (SNR) is a key parameter to consider for minimizing the bit error rate (BER). The inherent noise found in millimeter-wave systems is mainly a combination of white noise and phase noise. Increasing the SNR in wireless data transfer systems can lead to reliability and performance improvements. To address this issue, we propose to use a recurrent neural network (RNN) with a long short-term memory (LSTM) autoencoder architecture to achieve signal noise reduction. This design is based on a composite LSTM autoencoder with a single encoder layer and two decoder layers. A V-band receiver test bench is designed and fabricated to provide a high-speed wireless communication system. Constellation diagrams display the output signals measured for various random sequences of PSK and QAM modulated signals. The LSTM autoencoder is trained in real time using various noisy signals. The trained system is then used to reduce noise levels in the tested signals. The SNR of the designed receiver is of the order of 11.8dB, and it increases to 13.66dB using the three-level LSTM autoencoder. Consequently, the proposed algorithm reduces the bit error rate from 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−8</sup> to 10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−11</sup> . The performance of the proposed algorithm is comparable to other noise reduction strategies. Augmented denoised signals are fed into a ResNet-152 deep convolutional network to perform the final classification. The demodulation types are classified with an accuracy of 99.93%. This is confirmed by experimental measurements.

Detecting failed tethers in submerged floating tunnels using an LSTM autoencoder and DNN algorithms

Hierarchical clustering algorithm has low accuracy when processing high-dimensional data sets. In order to solve the problem, this paper presents a two-stage hierarchical clustering algorithm based on a Long Short Term Memory (LSTM) autoencoder. Firstly, we use LSTM autoencoder to learn the potential low-dimensional feature representation of the data. Secondly, we use the proposed two-stage hierarchical clustering algorithm to cluster the low-dimensional features. In the first stage, this algorithm divides the data points that are closest to each other into one cluster, for the data points have been clustered, their corresponding reverse nearest neighbors are also divided into the same cluster. In the second stage, the average density is defined as a measure, in each iteration, the cluster with the highest average density is merged with its nearest cluster until the threshold of the number of clusters is reached. Experiments on the UCI dataset show a significant improvement in the accuracy of the proposed algorithm when compared to the PERCH, BIRCH, CURE, SRC and RSRC algorithms.

Failure diagnosis on some system is often preferred even the data of the system is not designed for the condition monitoring and does not contain any or contains little example cases of failures. For this kind of system, it is unrealistic to directly observe condition from single feature or neither to build a machine learning system that has been trained to detect known failures. Still if any data describing the system exists, it is possible to provide some level of diagnosis on the system. Here we present an LSTM (Long Short Term Memory) autoencoder approach for detecting and isolating system failures with insufficient data conditions. Here we also illustrate how the failure isolation capability is effected by the choice of input feature space. The approach is tested with the flight data of F-18 aircraft and the applicability is validated against several leading edge flap (LEF) control surface seizure failures. The method shows a potential for not only detecting a potential failure in advance but also to isolate the failure by allocating the anomaly on the data to the features that are related to the operation of LEFs. The approach presented here provides diagnostic value from the data than is not designed for condition monitoring neither contain any example case failures.

A precise forecast of wind speed is a fundamental requirement of wind power integration. The nonlinear and intermittent nature of the wind makes wind speed forecasting (WSF) complicated for linear approaches. Addressing the complications faced by the linear approaches, this paper proposed a novel and robust approach using long short-term memory (LSTM) autoencoder, convolutional neural network (CNN), and LSTM model for enhanced WSF. The proposed hybrid approach is divided into two main components: feature encoding, dimensionality reduction using LSTM autoencoder and forecasting using convolutional LSTM. In the first stage, the LSTM autoencoder eliminates the uncertainties present in raw wind speed data and also reduces the computational load on the forecasting convolutional LSTM approach. Then, in the second stage, CNN is used to extract the optimum features, and the LSTM network is used to forecast the wind speed. Five different benchmark forecasting models are used to evaluate and study the proposed hybrid approach's performance. The experiment is performed with real-time wind speed data from the Garden city wind farm, USA. The proposed hybrid approach performance is verified using various performance metrics. The experimental results demonstrate that the proposed approach improved by 40% over the second best benchmark forecasting approach.

Automatic speech recognition systems (ASR) suffer from performance degradation under noisy conditions. Recent work, using deep neural networks to denoise spectral input features for robust ASR, have proved to be successful. In particular, Long Short-Term Memory (LSTM) autoencoders have outperformed other state of the art denoising systems when applied to the mfcc’s of a speech signal. In this paper we also consider denoising LSTM autoencoders (DLSTMA), but instead use three different DLSTMAs and apply each to the mfcc’s, fundamental frequency, and energy features, respectively. Results are given using several kinds of additive noise at different intensity levels, and show how this collection of DLSTMA’s improves the performance of the ASR in comparison with the LSTM autoencoder.

The safe operation of aero-engines is crucial for ensuring flight safety, and effective fault detection methods are fundamental to achieving this objective. In this paper, we propose a novel approach that integrates an auto-encoder with long short-term memory (LSTM) networks and a self-attention mechanism for the anomaly detection of aero-engine time-series data. The dataset utilized in this study was simulated from real data and injected with fault information. A fault detection model is developed utilizing normal data samples for training and faulty data samples for testing. The LSTM auto-encoder processes the time-series data through an encoder–decoder architecture, extracting latent representations and reconstructing the original inputs. Furthermore, the self-attention mechanism captures long-range dependencies and significant features within the sequences, thereby enhancing the detection accuracy of the model. Comparative analyses with the traditional LSTM auto-encoder, as well as one-class support vector machines (OC-SVM) and isolation forests (IF), reveal that the experimental results substantiate the feasibility and effectiveness of the proposed method, highlighting its potential value in engineering applications.

The sixth-generation (6G) wireless communication networks are anticipated to combine terrestrial, aerial, and marine communications into a dependable, fast network that could handle many devices with ultra-low latency requirements. Reconfigurable intelligent surfaces (RIS)-aided millimeter-wave massive MIMO communication systems could increase wireless link quality by providing passive beamforming gain via low-cost reflecting components. On the other hand, the power consumption and cost could be reduced much by applying a hybrid precoding architecture, which combines digital and analog precoding modules. However, using RIS to solve the problem of hybrid precoding is difficult because how to Figure out reflecting coefficients without beam training overhead or large channel estimation is a challenging issue. In this work, we propose a novel hybrid precoding architecture based on geometric mean decomposition (GMD) and jointly consider the design of the Long ShortTerm Memory (LSTM) autoencoder scheme. That is, we adopt GMD for diminishing the computational complexity and enhancing the hybrid precoding performance. Since the encoder-decoder design of autoencoder serves as a dimensionality reduction strategy, the LSTM autoencoder are able to capture the temporal and spatial distribution of the sequential data by using the LSTM models sequenhal and feature extraction capabilities. Numerical results show that our proposed algorithm could significantly improve the RIS-aided millimeter-wave MIMO communication systems performance compared with previous works.