Multi-sensor Time Series Research Articles

Classification of faults in friction stir processed composites using a machine learning and ensemble learning approach

Abstract Aluminium alloy based surface composites with hard reinforcement particles have wide scope in aerospace and automobile manufacturing industries. In this paper, the aluminium composites, manufactured by friction stir processing (FSP) with varying parameters are investigated for the faults occurred during fabrication process. It explores a machine-learning approach to detect defects of surface hybrid composites with an Al6061 alloy matrix, reinforced with copper and graphene nano-powders, using friction stir processing and a tungsten carbide tool on a milling machine. Multi-sensor time series data (vibration, force, and current) collected during fabrication, is preprocessed and labelled with normal and defective categories (e.g., pin break, brazing break, rough surface, no composite) using visual inspection. The important time domain and frequency domain features are extracted using different libraries in python. Thenafter, various types of feature selection techniques, viz filter, wrapper and embedded methods are implemented to select most relevant features. The selected subset of features from all selection methods used, are applied to different machine learning and ensemble learning classifiers and their performances are evaluated. The optimal combinations of the type of feature selection method and classifier used, are obtained for efficient classification of surface defects in composited formed by FSP. The real time monitoring and defect detection system can be developed in future for the composites developed by FSP using the developed models.

Machine learning approach for detection of land subsidence induced by underground coal fire using multi-sensor satellite data

ABSTRACT High-rank coal reserves in Jharia Coalfield (JCF, India), are invariably associated with underground coal fires and land subsidence. This study explores multi-sensor time series satellite data (Landsat 8 OLI and Sentinel-1) through machine learning (ML) to determine the regional ground deformation accompanying coal fires and their contextual relationship. The results show that the highest degree of subsidence is closely associated with the active mine benches with overburden dumps. The relationship between the coal fire and land subsidence parameters is considered as a binary classification problem, explored by calculating the probability of subsidence with a desirable categorical outcome through different ML models. The accuracy of the models is validated using performance metrics that shows that the Random Forest (RF) metrics predict the probability of deformation locations in response to the volume reduction of the burning coal fire and vertical compression due to Overburden Dump (OBD) near active mine benches. The estimated displacement trends have been used to forecast the Autoregressive Integrated Moving Average (ARIMA) method, estimated using Line-of-Sight (LOS) displacement values vary around the best fit within the 95% confidence limits. The trend shows ∼15–25% increase in subsidence compared to the cumulative subsidence.

TimeTector: A Twin-Branch Approach for Unsupervised Anomaly Detection in Livestock Sensor Noisy Data (TT-TBAD).

Unsupervised anomaly detection in multivariate time series sensor data is a complex task with diverse applications in different domains such as livestock farming and agriculture (LF&A), the Internet of Things (IoT), and human activity recognition (HAR). Advanced machine learning techniques are necessary to detect multi-sensor time series data anomalies. The primary focus of this research is to develop state-of-the-art machine learning methods for detecting anomalies in multi-sensor data. Time series sensors frequently produce multi-sensor data with anomalies, which makes it difficult to establish standard patterns that can capture spatial and temporal correlations. Our innovative approach enables the accurate identification of normal, abnormal, and noisy patterns, thus minimizing the risk of misinterpreting models when dealing with mixed noisy data during training. This can potentially result in the model deriving incorrect conclusions. To address these challenges, we propose a novel approach called "TimeTector-Twin-Branch Shared LSTM Autoencoder" which incorporates several Multi-Head Attention mechanisms. Additionally, our system now incorporates the Twin-Branch method which facilitates the simultaneous execution of multiple tasks, such as data reconstruction and prediction error, allowing for efficient multi-task learning. We also compare our proposed model to several benchmark anomaly detection models using our dataset, and the results show less error (MSE, MAE, and RMSE) in reconstruction and higher accuracy scores (precision, recall, and F1) against the baseline models, demonstrating that our approach outperforms these existing models.

Task-oriented analysis and visualization of correlation patterns in multi-sensor time series

Time series data are usually collected through instruments equipped with sensors. For multi-sensor time series (MSTS), it is crucial to identify the factors that affect tasks such as classification and regression. To better understand the mechanisms affecting the task, we extracted the correlation patterns between sensors and represented them as symmetric positive definite matrices. The correlation patterns were then transformed into vector form to serve as input features for regression or classification models, and the models were trained for each sensor. Finally, we leveraged the interpretability of the models to analyze and visualize the correlation patterns at both micro and macro scales. By integrating the explanatory power of the models with correlation patterns, we could interpret the task in terms of time and space, providing valuable insights for exploring the underlying data rules. We evaluated our proposed method using both synthetic and real data, and the simulation results confirmed its effectiveness.

Thermal Infrared Anomalies Associated with Recent Crustal Earthquakes in Gorj County in Romania

Cumulative stress-energy in tectonic active regions associated with ongoing deformation (co-seismic stress) along the main active geologic faults manifests various earthquakes’ precursors. Macro-fracturing processes are preceded by micro-fracturing phenomena which amplify strain field, and geophysical fields’ disturbances like space-time anomalies of Earth’s fracto-emission signals (Thermal; Ionospheric; Electric Field; Acoustic; Radon in air, underground water, and soil; Neutron, etc.) registered months to days before the occurrence of earthquakes. Multispectral and multisensor time series satellite data and space-based geodetic (GPS and GNSS) measurements in synergy with ground-based geophysical/geochemical measurements bring a significant contribution to survey pre-earthquake signals in the tectonic active areas. This study aims to assess the role of seismo-induced thermal infrared (TIR) anomalies characterized by the ground-level air-AT and land surface temperature-LST changes as significant precursors of crustal moderate and strong earthquakes recorded in Gorj County, Oltenia region in Romania. North-West Targu Jiu area is an active tectonic area characterized by two strike-slip faults systems of North-West- South West and East-West orientation, where on 13 February 2023 was recorded a shallow strong earthquake of moment magnitude Mw= 5.2 and 17 km focal depth, followed by 14 February 2023 stronger earthquake of Mw = 5.7 and 6 km focal depth. Several months beyond have been recorded more than 1000 aftershocks of Mw ≤ 4.9. Based on MODIS Terra/Aqua and NOAA AVHRR time series data, some days to weeks before the seismic sequence has detected LST and AT anomalies.

Data Fusion and Pattern Classification in Dynamical Systems Via Symbolic Time Series Analysis

Abstract Symbolic time series analysis (STSA) plays an important role in the investigation of continuously evolving dynamical systems, where the capability to interpret the joint effects of multiple sensor signals is essential for adequate representation of the embedded knowledge. This technical brief develops and validates, by simulation, an STSA-based algorithm to make timely decisions on dynamical systems for information fusion and pattern classification from ensembles of multisensor time series data. In this context, one of the most commonly used methods has been neural networks (NN) in their various configurations; however, these NN-based methods may require large-volume data and prolonged computational time for training. An alternative feasible method is the STSA-based probabilistic finite state automata (PFSA), which has been shown in recent literature to require significantly less training data and to be much faster than NN for training and, to some extent, for testing. This technical brief reports a modification of the current PFSA methods to accommodate (possibly heterogeneous and not necessarily tightly synchronized) multisensor data fusion and (supervised learning-based) pattern classification in real-time. Efficacy of the proposed method is demonstrated by fusion of time series of position and velocity sensor data, generated from a simulation model of the forced Duffing equation.

Assessing Combinations of Landsat, Sentinel-2 and Sentinel-1 Time series for Detecting Bark Beetle Infestations

ABSTRACT Bark beetle infestations are among the most substantial forest disturbance agents worldwide. Moreover, as a consequence of global climate change, they have increased in frequency and in the size and number of affected areas. Controlling bark beetle outbreaks requires consistent operational monitoring, as is possible using satellite data. However, while many satellite-based approaches have been developed, the full potential of dense, multi-sensor time series has yet to be fully explored. Here, for the first time, we used all available multispectral data from Landsat and Sentinel-2, Sentinel-1 SAR data, and combinations thereof to detect bark beetle infestations in the Bavarian Forest National Park. Based on a multi-year reference dataset of annual infested areas, we assessed the separability between healthy and infested forests for various vegetation indices calculated from the satellite data. We used two approaches to compute infestation probability time series from the different datasets: Bayesian conditional probabilities, based on the best-separating index from each satellite type, and random forest regression, based on all indices from each satellite type. Five different sensor configurations were tested for their detection capabilities: Landsat alone, Sentinel-1 alone, Sentinel-2 alone, Landsat and Sentinel-2 combined, and data from all satellite types combined. The best overall results in terms of spatial accuracy were achieved with Sentinel-2 (max. overall accuracy: 0.93). The detections of Sentinel-2 also were the closest to the onset of infestation estimated for each year. Sentinel-2 detected infested areas in larger contiguous patches with higher reliability compared to smaller patches. The results achieved with Landsat were somewhat inferior to those of Sentinel-2 (max. accuracy: 0.89). While yielding similar results, the combination of Landsat and Sentinel-2 did not provide any advantages over using Landsat or Sentinel-2 alone (max. accuracy: 0.87), while Sentinel-1 was unable to detect infested areas (max. accuracy: 0.62). The combined data of all three satellite types did not achieve satisfactory results either (max. accuracy: 0.67). Spatial accuracies were typically higher for Bayesian conditional probabilities than for random forest-derived probabilities, but the latter resulted in earlier detections. The approach presented herein provides a flexible disturbance detection pipeline well-suited for the monitoring of bark beetle outbreaks. Furthermore, it can also be applied to other disturbance types.

A decade-plus of Antarctic sea ice thickness and volume estimates from CryoSat-2 using a physical model and waveform fitting

Abstract. We estimate the snow depth and snow freeboard of Antarctic sea ice using a comprehensive retrieval method (referred to as CryoSat-2 Waveform Fitting for Antarctic sea ice, or CS2WFA) consisting of a physical waveform model and a waveform-fitting process that fits modeled waveforms to CryoSat-2 data. These snow depth and snow freeboard estimates are combined with snow, sea ice, and sea water density values to calculate the sea ice thickness and volume over an 11+ year span between 2010 and 2021. We first compare our snow freeboard, snow depth, and sea ice thickness estimates to other altimetry- and ship-based observations and find good agreement overall in both along-track and monthly gridded comparisons. Some discrepancies exist in certain regions and seasons that are theorized to come from both sampling biases and the differing assumptions in the retrieval methods. We then present an 11+ year time series of sea ice thickness and volume both regionally and pan-Antarctic. This time series is used to uncover intra-decadal changes in the ice cover between 2010 and 2021, showing small, competing regional thickness changes of less than 0.5 cm yr−1 in magnitude. Finally, we place these thickness estimates in the context of a longer-term, snow freeboard-derived, laser–radar sea ice thickness time series that began with NASA's Ice, Cloud, and land Elevation Satellite (ICESat) and continues with ICESat-2 and contend that reconciling and validating this longer-term, multi-sensor time series will be important in better understanding changes in the Antarctic sea ice cover.

A multimode structured prediction model based on dynamic attribution graph attention network for complex industrial processes

Complex industrial processes with dynamic and time-varying characteristics, as well as diverse operating conditions, pose challenges in developing accurate real-time online prediction models for the key performance indicator (KPI). To address these challenges, a KPI prediction framework is proposed in this study, which utilizes a multimode structured prediction model based on a dynamic-attribution graph attention network (DAGAT). First, a multimode clustering method based on a Gaussian mixture model is designed to divide and locate industrial multimode operational data. Second, to address the limitations of traditional deep network models in accurately capturing the complex relationships between process variables and KPI, a graph structure is used to model multi-sensor time series data. Subsequently, a novel DAGAT is proposed to improve node prediction performance and interpretation of industrial graph data. Considering the cumulative effect of large-scale equipment, the DAGAT incorporates knowledge of the industrial process, enabling it to capture the influence of chemical-reaction variables. Simultaneously, the Huber loss function is used to further optimize the prediction results. Moreover, the model can calculate in real-time the contribution of each individual process variable to KPI, providing guidance for optimal control. Finally, the effectiveness and superiority of the proposed method are validated through an industrial case.

Multi-Channel Time-Frequency Domain Deep CNN Approach for Machinery Fault Recognition Using Multi-Sensor Time-Series

Multi-Channel Time-Frequency Domain Deep CNN Approach for Machinery Fault Recognition Using Multi-Sensor Time-Series

Tensor Singular Spectrum Decomposition: Multisensor Denoising Algorithm and Application

Realising multi-sensor signal fusion and weak feature adaptive extraction is a challenging task. Therefore, a new algorithm called tensor singular spectrum decomposition is proposed in this study for the adaptive decomposition of multi-sensor time series. Traditional tensor decomposition algorithms, such as CP, HOSVD and Tucker decomposition, are derived from <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</i> -mode product. The <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</i> -mode product essentially uses the idea of matrices to deal with tensors given that it defines the multiplication between matrix and higher-order tensor, thereby creating problems of non-pseudodiagonal core tensor and nonunique decomposition results in traditional tensor decomposition algorithms. To this end, decomposition of the original tensor signal and reconstruction of multi-sensor component signals are realised in this study by combining the trajectory tensor construction, superposition of the Gaussian function spectral model, adaptive iterative optimisation of embedding dimension and diagonal average method on the basis of the principle of tensor–tensor order-preserving multiplication. The proposed algorithm inherits the perfect mathematical theory and excellent properties of matrix singular value decomposition in processing single-sensor signals whilst retaining the inherent structure and coupling relationship between multi-sensor data and realising the organic fusion and adaptive decomposition of multi-sensor signals. The analysis results of simulation, experimental and engineering signals showed that the proposed method can effectively extract weak fault quantification features hidden in original multi-sensor signals compared with existing methods.

Root cause diagnosis for process faults based on multisensor time-series causality discovery

Root cause diagnosis (RCD) focuses on locating the critical root causes and identifying the propagation paths of the industrial process faults, which is superior to conventional fault diagnosis methods and has attracted extensive attention. However, RCD is challenging in terms of dense sensor layout and diversified interconnected subsystems in complex industrial processes, since a fault originating in one unit can propagate throughout the entire plant along with the flow of information and material, obscuring its root causes. Moreover, the previous RCD methods ignore the more crucial temporal information between multisensor time-series, suffering from the disadvantages of requiring expert knowledge and responding untimely. Therefore, a novel data-driven RCD methodology named multisensor time-series causality discovery (MTCD) is proposed for industrial processes fault diagnosis. Firstly, a temporal registration network (TRN) based on the dilated convolutional neural network (DCNN) is proposed to extract the temporal correlation between time-series and implement the time-series prediction. Ulteriorly, a permutation importance causality validation (PICV) is designed to verify the causality between time-series based on the TRN prediction results without prior knowledge or expert experience. Besides, a causal time delay discovery approach based on layerwise relevance propagation (LRP) is presented by interpreting the weights of the trained TRN. Finally, we demonstrate the effectiveness of the proposed method with a stochastic synthesis process and the Tennessee Eastman process. The experimental results show that the proposed method has better performance in both root cause localization and propagation path identification.

A transformer‐based end‐to‐end data‐driven model for multisensor time series monitoring of machine tool condition

AbstractOnline determination of a cutter's health status is crucial for the attainment of condition‐based automated tool change in computer numerically controlled (CNC) machining. Due to the impracticalities associated with direct condition measurements, data‐based modeling of monitoring signals provides a viable practical route. However, the highly noisy and redundant nature of the associated data impacts negatively on model's accuracy and typically calls for additional initial preprocessing before modeling. Additionally, the long sequential data entails widely varying condition distributions exhibited by different cutters, even from the same batch on similar machining parameters, posing a challenge to model generalization. An end‐to‐end model has thus been developed to work directly on unprocessed data to establish global sensitive features from varying distributions for online tool wear estimation in CNC machining. The model utilizes three main functional blocks. First, a data denoising and feature selection block automatically processes raw multisensor data directly, dispensing with scaling or preprocessing of inputs as conventionally done. Each sensor channel's independence is preserved at initial processing ensuring complementary information from different sensors is utilized while simultaneously minimizing existing redundancies. The weighted denoised data is then processed through a transformer encoder block for determination of global dependencies in the time‐series sequence, regardless of the time‐step position. The learned features are then fed to an upper supervised learning block for association with the monitored wear condition. The developed model works directly on raw noisy data irrespective of scaling differences, saving on preprocessing computational cost. The global associations extracted on long sequences by the transformer‐encoder allow for model generalization to varying wear distributions. The parallel processing structure of all channels ensures complementary information is utilized minimizing unforeseen model bias. The model's performance as evaluated on experimental milling data and further comparison with other reported models on same dataset shows attainment of comparable state‐of‐art results.

A self-supervised algorithm to detect signs of social isolation in the elderly from daily activity sequences

Considering the increasing aging of the population, multi-device monitoring of the activities of daily living (ADL) of older people becomes crucial to support independent living and early detection of symptoms of mental illnesses, such as depression and Alzheimer’s disease. Anomalies can anticipate the diagnosis of these pathologies in the patient’s normal behavior, such as reduced hygiene, changes in sleep habits, and fewer social interactions. These abnormalities are often subtle and hard to detect. Especially using non-intrusive monitoring devices might cause anomaly detectors to generate false alarms or ignore relevant clues. This limitation may hinder their usage by caregivers. Furthermore, the notion of abnormality here is context and patient-dependent, thus requiring untrained approaches.To reduce these problems, we propose a self-supervised model for multi-sensor time series signals based on Hyperbolic uncertainty for Anomaly Detection, which we dub HypAD. HypAD estimates uncertainty end-to-end, thanks to hyperbolic neural networks, and integrates it into the ”classic” notion of reconstruction loss in anomaly detection. Based on hyperbolic uncertainty, HypAD introduces the principle of a detectable anomaly. HypAD assesses whether it is sure about the input signal and fails to reconstruct it because it is anomalous or whether the high reconstruction loss is due to the model uncertainty, e.g., a complex but regular signal (cf. this parallels the residual model error upon training).The proposed solution has been incorporated into an end-to-end ADL monitoring system for elderly patients in retirement homes, developed within a funded project leveraging an interdisciplinary consortium of computer scientists, engineers, and geriatricians. Healthcare professionals were involved in the design and verification process to foster trust in the system. In addition, the system has been equipped with explainability features.

Mapping Samudra Tapu glacier: A holistic approach utilizing radar and optical remote sensing data for glacier radar facies mapping and velocity estimation

Himalayan glaciers have shown more sensitivity and visible changes to the climate change and global warming in the last 150 years. The highly rugged topography and inaccessible remote areas makes satellite images as the most appropriate source of information retrieval. We performed remote sensing based glacier change study for Samudra Tapu glacier, located in the Chandra basin of North-West Himalaya. In the present study, the capabilities of both optical and microwave remote sensing data was analysed in glacier change study in terms of its coverage, shift in equilibrium line altitude (ELA) and surface velocity over a period from 2000 to 2021. Multi Sensor (RISAT-1, Sentinel-1) time series of C-band SAR data along with a object oriented classification technique were used to identify different glacier facies such as percolation facies, icefalls, bare ice facies, refreeze snow and supraglacial debris. These classified maps were also used to detect the snow line and firn line along with ELA, aided with elevation information from digital elevation model (DEM). It was identified that more than 50 % of the total glacier area still lies into accumulation region. Further, we estimated the glacier surface velocity using Differential Interferometric Synthetic Aperture Radar (DInSAR) technique using European Remote Sensing Satellite (ERS-1/2) tandem data of 1996. High value of coherence was observed from the SAR return signal for one-day temporal difference. A mean velocity of 17–24 cm/day was found for the months of March and May 1996, highest flow rates were seen in the high accumulation area located in the Eastern and Southern Aspect of glacier. Spatial analysis of velocity patterns with respect to slope and aspect show that high rates of flow was found in southern slopes and movement rates generally increase with increase in slope. Feature tracking approach was used to estimate the glacier flow for long term and seasonal basis using optical and SAR datasets (IRS-1C, 1D PAN, Landsat-7, 8 PAN, and TANDEM-x) during 1999–2020 period. The results suggest that glacier flow varies with season, i.e., high velocity during spring-summer season, as compared to late summer or winter and, the rate of ice flow changes over the years. The mean glacier velocity reduced to 49.5 m/year during 2013–2020 time, as compared to 67.67 m/year during 1999–2003 time. These results of reducing glacier velocity and changing snow line altitude indicates enhanced glacier’s melt rate and overall negative mass balance for Smudra tapu glacier.

Phytoplankton Bloom Dynamics in the Baltic Sea Using a Consistently Reprocessed Time Series of Multi-Sensor Reflectance and Novel Chlorophyll-a Retrievals

A relevant indicator for the eutrophication status in the Baltic Sea is the Chlorophyll-a concentration (Chl-a). Alas, ocean color remote sensing applications to estimate Chl-a in this brackish basin, characterized by large gradients in salinity and dissolved organic matter, are hampered by its optical complexity and atmospheric correction limits. This study presents Chl-a retrieval improvements for a fully reprocessed multi-sensor time series of remote-sensing reflectances (Rrs) at ~1 km spatial resolution for the Baltic Sea. A new ensemble scheme based on multilayer perceptron neural net (MLP) bio-optical algorithms has been implemented to this end. The study documents that this approach outperforms band-ratio algorithms when compared to in situ datasets, reducing the gross overestimates of Chl-a observed in the literature for this basin. The Rrs and Chl-a time series were then exploited for eutrophication monitoring, providing a quantitative description of spring and summer phytoplankton blooms in the Baltic Sea over 1998–2019. The analysis of the phytoplankton dynamics enabled the identification of the latitudinal variations in the spring bloom phenology across the basin, the early blooming in spring in the last two decades, and the description of the spatiotemporal coverage of summer cyanobacterial blooms in the central and southern Baltic Sea.

Health condition monitoring of machines based on long short-term memory convolutional autoencoder

Machine health assessment is crucial to prognostics and health management (PHM), which can increase reliability of machines while reducing operation cost. However, the data collected from machines in real-world cases are noised and high-dimensional, so that it is very difficult to detect the early defects of machines. Moreover, it is still a challenging problem to capture sequential information from multi-sensor signals for machine health assessment. This paper proposes a novel autoencoder (AE), called long short-term memory convolutional autoencoder (LSTMCAE), where LSTM and convolutional units are embedded in this specific network for feature learning from sensor signals based on unsupervised-learning. A long short-term memory (LSTM) unit in LSTMCAE is utilized to capture sequential information from multi-sensor time series data. A convolutional and deconvolutional unit is further embedded after the LSTM unit to filter noise and extract features corresponding to health state of machines. Residual learning is employed to reduce the training difficulty and improve the feature learning performance of LSTMCAE. Multivariate Gaussian distribution (MGD) is adopted to generate health index (HI) based on reconstruction errors of LSTMCAE for quantifying machines health state. A contribution analysis-based feature selection method is proposed to select effective features for machinery health assessment. Experimental results on turbofan engines demonstrate the effectiveness of LSTMCAE for machine health assessment. Compared with other unsupervised learning methods, the HIs generated by LSTMCAE show better tendency. LSTMCAE can detect earlier slight degradation than LSTM-AE. Moreover, contribution analysis results indicate that the sensitive variables can be selected for machine health monitoring. Residual learning significantly improves the feature learning performance of LSTMCAE. Thus, LSTMCAE can well quantify the degradation trend of machines. The experimental results on engines show that the proposed method will be an effective tool for machine health assessment.

Unsupervised Deep Anomaly Detection for Multi-Sensor Time-Series Signals

Nowadays, multi-sensor technologies are applied in many fields, e.g., Health Care (HC) and Human Activity Recognition (HAR). These sensors can generate a substantial amount of multivariate time-series data. Unsupervised anomaly detection on multi-sensor time-series data has been proven critical in machine learning researches. The key challenge is to discover generalized normal patterns by capturing spatial-temporal correlation in time series. Beyond this challenge, noisy data is often intertwined with training data, which is likely to mislead the model by making it hard to distinguish between normal, abnormal, and noisy data. Few of previous researches can jointly address these two challenges. In this paper, we propose a novel deep learning-based anomaly detection algorithm called Deep Convolutional Autoencoding Memory network (CAE-M). We first build a Deep Convolutional Autoencoder to characterize spatial dependence of high-dimensional data with a Maximum Mean Discrepancy (MMD) to better distinguish between the noisy, normal, and abnormal data. Then, we construct a Memory Network consisting of linear (Autoregressive Model) and non-linear predictions (Bidirectional LSTM with Attention) to capture temporal dependence from time-series data. Finally, CAE-M jointly optimizes these two subnetworks. We compare the proposed approach with several state-of-the-art anomaly detection methods on HAR and HC datasets. Experimental results demonstrate that our proposed model outperforms these existing method

Multisensor Prognostic of RUL Based on EMD-ESN

This paper presents a prognostic method for RUL (remaining useful life) prediction based on EMD (empirical mode decomposition)-ESN (echo state network). The combination method adopts EMD to decompose the multisensor time series into a bunch of IMFs (intrinsic mode functions), which are then predicted by ESNs, and the outputs of each ESN are summarized to obtain the final prediction value. The EMD can decompose the original data into simpler portions and during the decomposition process, much noise is filtered out and the subsequent prediction is much easier. The ESN is a relatively new type of RNN (recurrent neural network), which substitutes the hidden layers with a reservoir remaining unchanged during the training phase. The characteristic makes the training time of ESN is much shorter than traditional RNN. The proposed method is applied to the turbofan engine datasets and is compared with LSTM (Long Short-Term Memory) and ESN. Extensive experimental results show that the prediction performance and efficiency are much improved by the proposed method.

Local-scale agricultural drought monitoring with satellite-based multi-sensor time-series

ABSTRACT Globally, drought constitutes a serious threat to food and water security. The complexity and multivariate nature of drought challenges its assessment, especially at local scales. The study aimed to assess spatiotemporal patterns of crop condition and drought impact at the spatial scale of field management units with a combined use of time-series from optical (Landsat, MODIS, Sentinel-2) and Synthetic Aperture Radar (SAR) (Sentinel 1) data. Several indicators were derived such as Normalized Difference Vegetation Index (NDVI), Normalized Difference Moisture Index (NDMI), Land Surface Temperature (LST), Tasseled cap indices and Sentinel-1 based backscattering intensity and relative surface moisture. We used logistic regression to evaluate the drought-induced variability of remotely sensed parameters estimated for different phases of crop growth. The parameters with the highest prediction rate were further used to estimate thresholds for drought/non-drought classification. The models were evaluated using the area under the receiver operating characteristic curve and validated with in-situ data. The results revealed that not all remotely sensed variables respond in the same manner to drought conditions. Growing season maximum NDVI and NDMI (70–75%) and SAR derived metrics (60%) reflect specifically the impact of agricultural drought. These metrics also depict stress affected areas with a larger spatial extent. LST was a useful indicator of crop condition especially for maize and sunflower with prediction rates of 86% and 71%, respectively. The developed approach can be further used to assess crop condition and to support decision-making in areas which are more susceptible and vulnerable to drought.