Temporal Dependence Research Articles

Leveraging Hybrid Deep Learning Models for Enhanced Multivariate Time Series Forecasting

Time series forecasting is crucial in various domains, ranging from finance and economics to weather prediction and supply chain management. Traditional statistical methods and machine learning models have been widely used for this task. However, they often face limitations in capturing complex temporal dependencies and handling multivariate time series data. In recent years, deep learning models have emerged as a promising solution for overcoming these limitations. This paper investigates how deep learning, specifically hybrid models, can enhance time series forecasting and address the shortcomings of traditional approaches. This dual capability handles intricate variable interdependencies and non-stationarities in multivariate forecasting. Our results show that the hybrid models achieved lower error rates and higher R2\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R^2$$\\end{document} values, signifying their superior predictive performance and generalization capabilities. These architectures effectively extract spatial features and temporal dynamics in multivariate time series by combining convolutional and recurrent modules. This study evaluates deep learning models, specifically hybrid architectures, for multivariate time series forecasting. On two real-world datasets - Traffic Volume and Air Quality - the TCN-BiLSTM model achieved the best overall performance. For Traffic Volume, the TCN-BiLSTM model achieved an R2\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R^2$$\\end{document} score of 0.976, and for Air Quality, it reached an R2\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R^2$$\\end{document} score of 0.94. These results highlight the model’s effectiveness in leveraging the strengths of Temporal Convolutional Networks (TCNs) for capturing multi-scale temporal patterns and Bidirectional Long Short-Term Memory (BiLSTMs) for retaining contextual information, thereby enhancing the accuracy of time series forecasting.

A new data-driven framework for progressive anomaly event alerts in spacecraft based on reconstruction discrepancy

In the use of deep learning for spacecraft anomaly detection, a key issue arises from the insufficient extraction of temporal dependencies in telemetry data. This can lead to an inability to accurately discern whether distribution changes in the data are caused by substantive anomalies or merely a consequence of model underfitting. To address this issue, we design a Temporal Dependency Extraction Enhanced Autoencoder model for multi-scale learning of telemetry data. Firstly, this model incorporates Multi-Scale Temporal Dependency Extraction blocks, which integrate self-attention, autoregressive, and feed-forward networks, aimed at systematically dissecting the long-term dependencies, historical information, and complex patterns in telemetry data. Building on these blocks, our model can efficiently and accurately reconstruct telemetry data while maintaining computational efficiency. Furthermore, we utilize an anomaly quantification metric based on the smoothed Manhattan distance, combined with the Drift Streaming Peaks-over-Threshold strategy for setting anomaly thresholds, thus establishing a comprehensive and precise anomaly alerts framework. Finally, we validate our approach using a dataset from the Attitude Control System of a Geostationary Earth Orbit satellite. The experimental results show that our method not only detects anomalies earlier than traditional methods but also provides an in-depth quantitative analysis of anomaly characteristics.

TADGCN: A Time-Aware Dynamic Graph Convolution Network for long-term traffic flow prediction

Traffic flow prediction is a crucial component of the Intelligent Transportation System (ITS) with substantial implications for traffic management and daily travel. While short-term traffic forecasting has shown promise, long-term prediction presents unique challenges. The existing short-term approaches fall short in long-term scenarios due to two primary issues: (1) ignoring modeling the interaction relationships among dynamic spatial dependencies, and (2) extracting long-term temporal dependencies without capturing the time information in-depth. To mitigate the above issues, we propose a novel model called TADGCN, a Time-Aware Dynamic Graph Convolution Network, to effectively perform long-term traffic flow prediction. TADGCN captures latent interactions among dynamic spatial dependencies using Attention-Driven Dynamic Adaptive Graph Convolution Network (ADAGCN) modules. The time-aware capability is effectively enhanced using Time-Aware Joint Multi-View Temporal Encoder (TJMTE) modules due to the construction of a time-aware matrix. We conduct diversified experiments on three real-world datasets (England, PEMS04, and PEMS08) to evaluate the performance of our proposed method TADGCN. The results reveal that TADGCN achieves the best accuracy in all cases against 16 competitive state-of-the-art baselines.

Attention mechanism enhanced LSTM networks for latency prediction in deterministic MEC networks

In deterministic mobile edge computing (MEC) networks, accurately predicting latency is critical for optimizing resource allocation and enhancing quality of service (QoS). This paper introduces a novel approach leveraging attention mechanism enhanced long short-term memory (LSTM) networks to predict latency in MEC networks. The proposed model integrates attention mechanisms into LSTM networks to capture temporal dependency and emphasize relevant features in the input data, thereby improving the prediction accuracy. T extensive experiments are conducted by using practical MEC network data, demonstrating that the proposed approach significantly outperforms traditional LSTM and other baseline models in terms of prediction accuracy and computational efficiency. Additionally, we analyze the impact of various configurations in the attention mechanism and LSTM on the model performance, providing insights into the optimal settings. The findings of this study contribute to the advancement of latency prediction techniques in deterministic MEC networks, facilitating more efficient and reliable network management.

A novel fault early warning method for centrifugal blowers based on stacked denoising autoencoder and transfer learning

Centrifugal blowers are easy to get faults due to the harsh working environment, and appropriate fault early warning is of great significance for predictive maintenance. Traditional fault early warning methods have poor resistance and feature learning ability in dealing with multivariate data with noise, and cannot achieve domain adaptation in different working environments. Aimed at solving these problems, this paper proposes a novel fault early warning method for centrifugal blowers based on stacked denoising autoencoder with sliding window (SW-SDAE) and transfer learning. The developed SW-SDAE model can effectively learn representative degradation features and temporal dependence from multivariate time-series data with noise. The reconstruction errors of SW-SDAE are used to construct the health indicators, which accurately characterizes the health status of the centrifugal blower. Meanwhile, transfer learning is employed to solve the problem of domain adaptation for different working environments. The established source domain warning model is successfully transferred to the target domain by minimizing the maximum mean discrepancy. When the health indicator exceeds the warning threshold, a fault early warning is performed. Experimental results demonstrate that the developed SW-SDAE warning model integrating transfer learning significantly resists the interference of noise and improves the domain adaptability for different working conditions. The proposed method achieves fault early warning 5.67 h without false alarms before failure and shows superior warning performance compared with traditional warning methods.

The improved integrated Exponential Smoothing based CNN-LSTM algorithm to forecast the day ahead electricity price

The deregulation of electricity market has led to the development of the short-term electricity market. The power generators and consumers can sell and purchase the electricity in the day ahead terms. The market clearing electricity price varies throughout the day due to the increase in the consumers bidding for electricity. Forecasting of the electricity in the day ahead market is of significance for appropriate bidding. To predict the electricity price the modified method of Exponential Smoothing-CNN-LSTM is proposed based on the time series method of Exponential Smoothing and Deep Learning methods of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM). The dataset used for assessment of the forecasting algorithms is collected from the day ahead electricity market at the Indian Energy Exchange (IEX). The forecasting results of the Exponential Smoothing-CNN-LSTM method evaluated in terms of Mean Absolute Error (MAE) as 0.11, Root Mean Squared Error (RMSE) as 0.17 and Mean Absolute Percentage Error (MAPE) as 1.53 % indicates improved performance. The proposed algorithm can be used to forecast the time series in other domains as finance, retail, healthcare, manufacturing.•The method of Exponential Smoothing-CNN-LSTM is proposed for forecasting the electricity price a day ahead for accurate bidding for the short-term electricity market participants.•The forecasting results indicate the better performance of the proposed method than the existing techniques of Exponential Smoothing, LSTM and CNN-LSTM due to the advantages of the Exponential Smoothing to extract the levels and seasonality and with the CNN-LSTM methods ability to model the complex spatial and temporal dependencies in the time series.

Patch ModernTCN-Mixer: a dual-task temporal convolutional network framework for hybrid implementation of first prediction time detection and remaining useful life prognosis

Abstract Over the past few years, notable advancements have been achieved in predicting the remaining useful life (RUL) of rotating equipment through deep learning methodologies. However, existing RUL prediction models tend to implement the determination of the first prediction time (FPT) for stage division separately from the RUL prediction, ignoring their potential correlation in the degradation process. In response to this issue, this paper proposed a dual-task prediction network framework based on Patch ModernTCN-Mixer (PMTCN-Mixer), which adaptively and jointly achieved FPT detection and RUL prediction. Firstly, the network designed a hard sharing parameter feature extractor module, Patch ModernTCN, which is used to learn the temporal dependence and spatial correlation of degradation features. Secondly, to eliminate redundant information and noise during the feature extraction phase while enhancing the precision of detection and prognosis, a dynamic semi-soft thresholding (DST) module was constructed. Lastly, the dual-task learning network PMTCN-Mixer was constructed by combining Patch ModernTCN with DST, utilizing GradNorm to balance the gradients between FPT detection and RUL prediction tasks to achieve fusion prediction. The performance of the PMTCN-Mixer framework was validated on the XJTU-SY Bearing Datasets and IEEE PHM 2012 Challenge Datasets, compared with the state-of-the-art network’s optimal results, the RUL prediction metrics root mean square error, mean absolute error, Score and R 2 were improved by 17.31%, 23.59%, 10.66%, and 4.76%, respectively. The findings confirm that the dual-task prediction model PMTCN-Mixer effectively captures both spatial and temporal semantic information from deterioration data, precisely accomplishes the integration of FPT detection and RUL prediction, and possess good generalization ability and superiority.

Railways Tender Price Prediction Using Machine Learning and Deep Learning Algorithms

The idea of determining the optimum bid amount in any railway tender competition has been a complex and critical task. Reasonable pricing is the main gateway in the process of winning contracts. The approach presented here is based on data-driven pricing to maximize tender wins and reduce potential profit loss. The model shall make use of historical tender data, competitor pricing data, and market indicators in order to predict the optimal bid amounts. This advanced ML model runs advanced machine learning algorithms, including regression models and ensembles, to study the intricate relationship between factors that would affect the successful execution of a bid. Deep learning models are integrated into the model to provide it with better handling of temporal dependencies and other hidden patterns in data, hence yielding accurate and robust predictions. The key objective of the work is to increase the winning rate of tender contracts by at least 10%, while competitive profitability is ensured. Because of precise competitor price predictions, business success criteria are oriented to reaching this higher win rate on tenders. The Machine learning success criteria target a price prediction accuracy of at least 90%. Another important set of economic success criteria that should be targeted is an improvement in profit margins of at least 5% through more accurate pricing strategies and a reduction in the number of rejected bids. It will offer great value to the businesses operating in the railway industry in making proper decisions on operations and strategic planning. The paper develops a fusion of traditional statistical methodologies with advanced ML and DL techniques in order to provide a robust solution for competitive advantage and increased profitability in the dynamic and competitive railway tender market.

Crested Porcupine Optimizer-Optimized CNN-BiLSTM-Attention Model for Predicting Main Girder Temperature in Bridges

Stage-built long-span bridges deform with temperature, affecting alignment to design needs. In this paper, a model for predicting temperature time series is proposed, which can predict temperatures in engineering practice and utilize the predicted results to adjust the elevation of stage construction. The model employs convolutional neural networks (CNNs) for initial feature extraction, followed by bidirectional long short-term memory (BiLSTM) layers to capture temporal dependencies. An attention mechanism is applied to the LSTM output, enhancing the model’s ability to focus on the most relevant parts of the sequence. The Crested Porcupine Optimizer (CPO) is used to fine-tune parameters like the number of LSTM units, dropout rate, and learning rate. The experiments on the measured temperature data of an under-construction cable-stayed bridge are conducted to validate our model. The results indicate that our model outperforms the other five models in comparison, with all the R2 values exceeding 0.97. The average of the mean absolute error (MAE) on the 30 measure points is 0.19095, and the average of the root mean square error (RMSE) is 0.28283. Furthermore, the model’s low sensitivity to data makes it adaptable and effective for predicting temperatures and adjusting the elevation in large-span bridge construction.

MCGnet with multi-level gated unit transformations: A time series prediction-based model for unknown pattern abnormality detection

The task of anomaly detection in data patterns remains challenging due to the diverse of fault patterns, which often render existing models ineffective when encountered with unknown or unseen data patterns in the training data. Furthermore, insufficient samples, the presence of chaotic behavior, unbalanced anomaly patterns are potential limiting factors in the practical application. While prediction error-based approaches attempt to address these challenges, the discriminability of abnormal patterns in existing scenarios is not yet sufficient, thus limiting the ability to identify abnormal phenomena. To overcome these limitations, we proposed a novel model called MCGnet. The backbone architecture of MCGnet employed a novel design of continuous multi-layer circular convolution transformations. This design enhances the expressive capability of the model by expanding the differences among data patterns. Additionally, we introduced in parallel the newly proposed temporal-frequency gated Gaussian unit, which generates richer feature representations. Subsequently, the obtained feature representations were fed into an explicit causal network to learn complex patterns of data along temporal dependencies. The MCGnet not only solves high-precision time series prediction problems but also significantly improves the discriminability between normal and abnormal patterns. Since the proposed model is accurately modeled using only a small amount of normal data, it bypasses the limiting factors such as insufficient fault samples and sample imbalance, and is able to effectively deal with various unknown data patterns. The accuracy of the proposed method was verified through simulation data (Duffing system) and its potential for engineering applications was demonstrated through real-world data (rotor system).

A Novel Deep Learning Approach for Real-Time Critical Assessment in Smart Urban Infrastructure Systems

The swift advancement of communication and information technologies has transformed urban infrastructures into smart cities. Traditional assessment methods face challenges in capturing the complex interdependencies and temporal dynamics inherent in these systems, risking urban resilience. This study aims to enhance the criticality assessment of geographic zones within smart cities by introducing a novel deep learning architecture. Utilizing Convolutional Neural Networks (CNNs) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal dependency modeling, the proposed framework processes inputs such as total electricity use, flooding levels, population, poverty rates, and energy consumption. The CNN component constructs hierarchical feature maps through successive convolution and pooling operations, while the LSTM captures sequence-based patterns. Fully connected layers integrate these features to generate final predictions. Implemented in Python using TensorFlow and Keras on an Intel Core i7 system with 32 GB RAM and an NVIDIA GTX 1080 Ti GPU, the model demonstrated a superior performance. It achieved a mean absolute error of 0.042, root mean square error of 0.067, and an R-squared value of 0.935, outperforming existing methodologies in real-time adaptability and resource efficiency.

Revolutionising waste management with the impact of Long Short-Term Memory networks on recycling rate predictions

This study explores the efficacy of Long Short-Term Memory (LSTM) networks in predicting recycling rates and enhancing resource allocation in waste management systems. It addresses the limitations of traditional statistical models and machine learning algorithms that struggle with sequential data and temporal dependencies. The methodology comprised collecting extensive datasets from public repositories, configuring the LSTM network architecture, training the model with historical data, and testing various activation functions and hyperparameters. The model’s performance was rigorously compared to traditional models and alternative machine learning algorithms using metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared (R2). The findings demonstrate that the LSTM model significantly outperformed traditional approaches, achieving an MAE of 3.5%, an RMSE of 2.8%, and an R2 of 0.92. These results underscore the superior capability of LSTM networks to capture complex temporal patterns in recycling data, offering substantial improvements in predictive accuracy and reliability. Consequently, the study highlights the potential of LSTM networks to revolutionize waste management strategies, contributing to more effective and sustainable practices.

New Insights on the Information Content of the Normalized Difference Vegetation Index Sentinel-2 Time Series for Assessing Vegetation Dynamics

The Sentinel-2 NDVI time series information content from 2017 to 2023 at a 10 m spatial resolution was evaluated based on the NDVI temporal dependency in five scenarios in central Spain. First, time series were interpolated and then filtered using the Savitzky–Golay, Fast Fourier Transform, Whittaker, and Maximum Value filters. Temporal dependency was assessed using the Q-Ljung-Box and Fisher’s Kappa tests, and similarity between raw and filtered time series was assessed using Correlation Coefficient and Root Mean Square Error. An Interpolating Efficiency Indicator (IEI) was proposed to summarize the number and temporal distribution of low-quality observations. Type of climate, atmospheric disturbances, land cover dynamics, and management were the main sources of variability in five scenarios: (1) rainfed wheat and barley presented high short-term variability due to clouds (lower IEI in winter and spring) during the growing cycle and high interannual variability due to precipitation; (2) maize showed stable summer cycles (high IEI) and low interannual variability due to irrigation; (3) irrigated alfalfa was cut five to six times during summer, resulting in specific intra-annual variability; (4) beech forest showed a strong and stable summer cycle, despite the short-term variability due to clouds (low IEI); and (5) evergreen pine forest had a highly variable growing cycle due to fast responses to temperature and precipitation through the year and medium IEI values. Interpolation after removing non-valid observations resulted in an increase in temporal dependency (Q-test), particularly a short term in areas with low IEI values. The information improvement made it possible to identify hidden periodicities and trends using the Fisher’s Kappa test. The SG filter showed high similarity values and weak influence on dynamics, while the MVF showed an overestimation of the NDVI values.

Attention-based CNN-BiLSTM for sleep state classification of spatiotemporal wide-field calcium imaging data

BackgroundWide-field calcium imaging (WFCI) with genetically encoded calcium indicators allows for spatiotemporal recordings of neuronal activity in mice. When applied to the study of sleep, WFCI data are manually scored into the sleep states of wakefulness, non-REM (NREM) and REM by use of adjunct EEG and EMG recordings. However, this process is time-consuming, invasive and often suffers from low inter- and intra-rater reliability. Therefore, an automated sleep state classification method that operates on spatiotemporal WFCI data is desired. New methodA hybrid network architecture consisting of a convolutional neural network (CNN) to extract spatial features of image frames and a bidirectional long short-term memory network (BiLSTM) with attention mechanism to identify temporal dependencies among different time points was proposed to classify WFCI data into states of wakefulness, NREM and REM sleep. ResultsSleep states were classified with an accuracy of 84 % and Cohen’s κ of 0.64. Gradient-weighted class activation maps revealed that the frontal region of the cortex carries more importance when classifying WFCI data into NREM sleep while posterior area contributes most to the identification of wakefulness. The attention scores indicated that the proposed network focuses on short- and long-range temporal dependency in a state-specific manner. Comparison with existing methodOn a held out, repeated 3-hour WFCI recording, the CNN-BiLSTM achieved a κ of 0.67, comparable to a κ of 0.65 corresponding to the human EEG/EMG-based scoring. ConclusionsThe CNN-BiLSTM effectively classifies sleep states from spatiotemporal WFCI data and will enable broader application of WFCI in sleep research.

Emotion recognition using hierarchical spatial–temporal learning transformer from regional to global brain

Emotion recognition is an essential but challenging task in human–computer interaction systems due to the distinctive spatial structures and dynamic temporal dependencies associated with each emotion. However, current approaches fail to accurately capture the intricate effects of electroencephalogram (EEG) signals across different brain regions on emotion recognition. Therefore, this paper designs a transformer-based method, denoted by R2G-STLT, which relies on a spatial–temporal transformer encoder with regional to global hierarchical learning that learns the representative spatiotemporal features from the electrode level to the brain-region level. The regional spatial–temporal transformer (RST-Trans) encoder is designed to obtain spatial information and context dependence at the electrode level aiming to learn the regional spatiotemporal features. Then, the global spatial–temporal transformer (GST-Trans) encoder is utilized to extract reliable global spatiotemporal features, reflecting the impact of various brain regions on emotion recognition tasks. Moreover, the multi-head attention mechanism is placed into the GST-Trans encoder to empower it to capture the long-range spatial–temporal information among the brain regions. Finally, subject-independent experiments are conducted on each frequency band of the DEAP, SEED, and SEED-IV datasets to assess the performance of the proposed model. Results indicate that the R2G-STLT model surpasses several state-of-the-art approaches.

Deep transfer learning-based bird species classification using mel spectrogram images.

The classification of bird species is of significant importance in the field of ornithology, as it plays an important role in assessing and monitoring environmental dynamics, including habitat modifications, migratory behaviors, levels of pollution, and disease occurrences. Traditional methods of bird classification, such as visual identification, were time-intensive and required a high level of expertise. However, audio-based bird species classification is a promising approach that can be used to automate bird species identification. This study aims to establish an audio-based bird species classification system for 264 Eastern African bird species employing modified deep transfer learning. In particular, the pre-trained EfficientNet technique was utilized for the investigation. The study adapts the fine-tune model to learn the pertinent patterns from mel spectrogram images specific to this bird species classification task. The fine-tuned EfficientNet model combined with a type of Recurrent Neural Networks (RNNs) namely Gated Recurrent Unit (GRU) and Long short-term memory (LSTM). RNNs are employed to capture the temporal dependencies in audio signals, thereby enhancing bird species classification accuracy. The dataset utilized in this work contains nearly 17,000 bird sound recordings across a diverse range of species. The experiment was conducted with several combinations of EfficientNet and RNNs, and EfficientNet-B7 with GRU surpasses other experimental models with an accuracy of 84.03% and a macro-average precision score of 0.8342.

Multi-attention network with redundant information filtering for multi-horizon forecasting in multivariate time series

Multi-horizon forecasting of multivariate time series has always been a prominent research topic in domains such as finance and transportation. While prediction models that integrate attention mechanisms and Seq2Seq frameworks effectively capture the temporal dependencies between sequences, there might exist distinctions in the contributions of different input variables to the predictions. RNN-based encoders often overlook these distinctions. Hence, we propose a Multi-attention Network with Redundant Information Filtering (MNRIF) for Multi-horizon Forecasting in Multivariate Time Series, comprising the underlying component Two-stage Feature Fusion Network (TSFFN) and the upper-level structure Temporal Attention Network(TAN). TSFFN first designs a variable-level attention to screen the spatial semantics of explanatory variables during multi-step recursive iterations, effectively reducing noisy information, then builds a feature gate fusion module to adaptively weigh the degree of association between prediction targets and spatial semantics as well as target variables, thereby accomplishing secondary filtering of redundant information. TAN is a task-oriented learner that combines the Encoder–Decoder framework with attention mechanism. By means of dual information filtering based on attention mechanism and recurrent gate, it enhances effective mining of long-term temporal dependency features in sequences. Extensive experiments conducted on three real-world benchmark datasets demonstrate MNRIF’s superior performance compared to various state-of-the-art methods of different types.

Self-attention with temporal prior: can we learn more from the arrow of time?

Many diverse phenomena in nature often inherently encode both short- and long-term temporal dependencies, which especially result from the direction of the flow of time. In this respect, we discovered experimental evidence suggesting that interrelations of these events are higher for closer time stamps. However, to be able for attention-based models to learn these regularities in short-term dependencies, it requires large amounts of data, which are often infeasible. This is because, while they are good at learning piece-wise temporal dependencies, attention-based models lack structures that encode biases in time series. As a resolution, we propose a simple and efficient method that enables attention layers to better encode the short-term temporal bias of these data sets by applying learnable, adaptive kernels directly to the attention matrices. We chose various prediction tasks for the experiments using Electronic Health Records (EHR) data sets since they are great examples with underlying long- and short-term temporal dependencies. Our experiments show exceptional classification results compared to best-performing models on most tasks and data sets.

Elevation-dependent biases of raw and bias-adjusted EURO-CORDEX regional climate models in the European Alps

Data from the EURO-CORDEX ensemble of regional climate model simulations and the CORDEX-Adjust dataset were evaluated over the European Alps using multiple gridded observational datasets. Biases, which are here defined as the difference between models and observations, were assessed as a function of the elevation for different climate indices that span average and extreme conditions. Moreover, we assessed the impact of different observational datasets on the evaluation, including E-OBS, APGD, and high-resolution national datasets. Furthermore, we assessed the bi-variate dependency of temperature and precipitation biases, their temporal evolution, and the impact of different bias adjustment methods and bias adjustment reference datasets. Biases in seasonal temperature, seasonal precipitation, and wet-day frequency were found to increase with elevation. Differences in temporal trends between RCMs and observations caused a temporal dependency of biases, which could be removed by detrending both observations and RCMs. The choice of the reference observation datasets used for bias adjustment turned out to be more relevant than the choice of the bias adjustment method itself. Consequently, climate change assessments in mountain regions need to pay particular attention to the choice of observational dataset and, furthermore, to the elevation dependence of biases and the increasing observational uncertainty with elevation in order to provide robust information on future climate.

Photodiode-based porosity prediction in laser powder bed fusion considering inter-hatch and inter-layer effects

Laser powder bed fusion, while promising, faces hurdles in certifying fabricated parts due to cost and complexity, with in-process monitoring emerging as a potential solution. Existing models focus on predicting defects at a given location using the monitoring signals from solely that same location. Hence, these models treat each track or layer independently of the previous and subsequent ones, neglecting potential interdependencies. This study proposed an in-situ, photodiode-based monitoring approach considering inter-hatch and inter-layer effects on porosity formation - factors often overlooked in existing research. Two Ti-6Al-4 V cuboids (10×10×5 mm3) were built with optimized process parameters, with the melt pool continuously monitored at 20 kHz via a co-axially mounted photodiode. The monitoring system captured the integral radiation in the near-infrared spectrum within a field of view centered on the melt pool. The porosity is assessed by X-ray computed tomography (X-CT), serving as ground truth to build supervised machine learning (ML) models. This study considered physical phenomena occurring during the printing process, including remelting of lack of fusion pores by the subsequent layer, keyholes penetrating the current layer hence introducing pores in the layer below, and overlap between adjacent scan tracks. These considerations are critical for a holistic understanding of pore formation mechanisms. Photodiode signals and computed tomography volumes were cropped using windows of four sizes to test the model's pore localization capability. A machine learning model, specifically a Convolutional Neural Network (CNN) - Long Short-Term Memory (LSTM) network, was trained to predict porosities using these window sequences. The CNN extracted spatial features from photodiode signals, addressing inter-hatch effects, while the LSTM captured temporal dependencies across layers, addressing inter-layer effects. The results, with the Area Under the Receiver Operating Characteristic curve (AUC) of 0.91 for pores exceeding 8000 μm3 in volume and 100 μm2 in cross-sectional area, demonstrate the feasibility of the proposed model in detecting pores-level defects. This high defect prediction and positioning accuracy are essential for process control, providing real-time status of the region of interest and informing the controller of pore positions, thus facilitating intra-layer or inter-layer correction.