Engineering Applications of Artificial Intelligence最新文献_第3页

A dual-method approach using autoencoders and transductive learning for remaining useful life estimation

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-24 DOI: 10.1016/j.engappai.2025.110285

Jing Yang , Nika Anoosha Boroojeni , Mehran Kazemi Chahardeh , Lip Yee Por , Roohallah Alizadehsani , U. Rajendra Acharya

Estimating the remaining useful life (RUL) of lithium-ion batteries presents a critical challenge, as it necessitates predicting their future performance and lifespan under diverse operational conditions. Addressing this issue is crucial for enhancing battery maintenance, improving reliability, and safeguarding devices that depend on lithium-ion technology. In this article, we propose a dual-method approach for RUL estimation. Firstly, an autoencoder (AE) extracts pivotal features from the input. Key measurable parameters, such as voltage, current, and temperature from charging profiles, are derived from the battery management system, providing robust data for the AE. The core of the AE is constructed using a spatial attention-based transductive long short-term memory (TLSTM) model, which is trained with an advanced generative adversarial network (GAN). The TLSTM model employs transductive learning, emphasizing samples near the test point to refine the fitting process and surpassing conventional LSTM models in performance. Following the AE training phase, the input's latent representation is inputted into a multilayer perceptron (MLP) designed for RUL prediction. We conduct thorough evaluations using National Aeronautics and Space Administration (NASA) datasets. Additionally, experiments from the Center for Advanced Life Cycle Engineering (CALCE) at the University of Maryland are underway to examine the influence of transfer learning (TL) on our model. The TLSTM model performs better than other deep learning models, achieving an impressive mean absolute percentage error (MAPE) ranging between 0.0053 and 0.0095. This highlights the efficacy and superiority of our approach in accurately predicting RUL, offering significant potential benefits for industries reliant on energy storage systems.

估算锂离子电池的剩余使用寿命（RUL）是一项严峻的挑战，因为这需要预测其在不同运行条件下的未来性能和寿命。解决这一问题对于加强电池维护、提高可靠性和保护依赖锂离子技术的设备至关重要。在本文中，我们提出了一种估算 RUL 的双重方法。首先，自动编码器（AE）从输入中提取关键特征。关键的可测量参数，如充电曲线中的电压、电流和温度，来自电池管理系统，为自动编码器提供可靠的数据。AE 的核心是使用基于空间注意力的传导式长短期记忆 (TLSTM) 模型构建的，该模型由先进的生成式对抗网络 (GAN) 进行训练。TLSTM 模型采用传导式学习，强调测试点附近的样本以完善拟合过程，在性能上超越了传统的 LSTM 模型。在 AE 训练阶段之后，输入的潜在表示被输入到为 RUL 预测而设计的多层感知器 (MLP)。我们使用美国国家航空航天局（NASA）的数据集进行了全面评估。此外，马里兰大学先进生命周期工程中心（CALCE）正在进行实验，以检验迁移学习（TL）对我们模型的影响。TLSTM 模型的表现优于其他深度学习模型，其平均绝对百分比误差（MAPE）介于 0.0053 和 0.0095 之间，令人印象深刻。这凸显了我们的方法在准确预测 RUL 方面的有效性和优越性，为依赖于储能系统的行业提供了巨大的潜在利益。

{"title":"A dual-method approach using autoencoders and transductive learning for remaining useful life estimation","authors":"Jing Yang , Nika Anoosha Boroojeni , Mehran Kazemi Chahardeh , Lip Yee Por , Roohallah Alizadehsani , U. Rajendra Acharya","doi":"10.1016/j.engappai.2025.110285","DOIUrl":"10.1016/j.engappai.2025.110285","url":null,"abstract":"<div><div>Estimating the remaining useful life (RUL) of lithium-ion batteries presents a critical challenge, as it necessitates predicting their future performance and lifespan under diverse operational conditions. Addressing this issue is crucial for enhancing battery maintenance, improving reliability, and safeguarding devices that depend on lithium-ion technology. In this article, we propose a dual-method approach for RUL estimation. Firstly, an autoencoder (AE) extracts pivotal features from the input. Key measurable parameters, such as voltage, current, and temperature from charging profiles, are derived from the battery management system, providing robust data for the AE. The core of the AE is constructed using a spatial attention-based transductive long short-term memory (TLSTM) model, which is trained with an advanced generative adversarial network (GAN). The TLSTM model employs transductive learning, emphasizing samples near the test point to refine the fitting process and surpassing conventional LSTM models in performance. Following the AE training phase, the input's latent representation is inputted into a multilayer perceptron (MLP) designed for RUL prediction. We conduct thorough evaluations using National Aeronautics and Space Administration (NASA) datasets. Additionally, experiments from the Center for Advanced Life Cycle Engineering (CALCE) at the University of Maryland are underway to examine the influence of transfer learning (TL) on our model. The TLSTM model performs better than other deep learning models, achieving an impressive mean absolute percentage error (MAPE) ranging between 0.0053 and 0.0095. This highlights the efficacy and superiority of our approach in accurately predicting RUL, offering significant potential benefits for industries reliant on energy storage systems.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"147 ","pages":"Article 110285"},"PeriodicalIF":7.5,"publicationDate":"2025-02-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143478994","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Explainable Differential Privacy-Hyperdimensional Computing for Balancing Privacy and Transparency in Additive Manufacturing Monitoring

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-23 DOI: 10.1016/j.engappai.2025.110282

Fardin Jalil Piran , Prathyush P. Poduval , Hamza Errahmouni Barkam , Mohsen Imani , Farhad Imani

Machine Learning (ML) models integrated with in-situ sensing offer transformative solutions for defect detection in Additive Manufacturing (AM), but this integration brings critical challenges in safeguarding sensitive data, such as part designs and material compositions. Differential Privacy (DP), which introduces mathematically controlled noise, provides a balance between data utility and privacy. However, black-box Artificial Intelligence (AI) models often obscure how this noise impacts model accuracy, complicating the optimization of privacy–accuracy trade-offs. This study introduces the Differential Privacy-Hyperdimensional Computing (DP-HD) framework, a novel approach combining Explainable AI (XAI) and vector symbolic paradigms to quantify and predict noise effects on accuracy using a Signal-to-Noise Ratio (SNR) metric. DP-HD enables precise tuning of DP noise levels, ensuring an optimal balance between privacy and performance. The framework has been validated using real-world AM data, demonstrating its applicability to industrial environments. Experimental results demonstrate DP-HD’s capability to achieve state-of-the-art accuracy (94.43%) with robust privacy protections in anomaly detection for AM, even under significant noise conditions. Beyond AM, DP-HD holds substantial promise for broader applications in privacy-sensitive domains such as healthcare, financial services, and government data management, where securing sensitive data while maintaining high ML performance is paramount.

{"title":"Explainable Differential Privacy-Hyperdimensional Computing for Balancing Privacy and Transparency in Additive Manufacturing Monitoring","authors":"Fardin Jalil Piran , Prathyush P. Poduval , Hamza Errahmouni Barkam , Mohsen Imani , Farhad Imani","doi":"10.1016/j.engappai.2025.110282","DOIUrl":"10.1016/j.engappai.2025.110282","url":null,"abstract":"<div><div>Machine Learning (ML) models integrated with in-situ sensing offer transformative solutions for defect detection in Additive Manufacturing (AM), but this integration brings critical challenges in safeguarding sensitive data, such as part designs and material compositions. Differential Privacy (DP), which introduces mathematically controlled noise, provides a balance between data utility and privacy. However, black-box Artificial Intelligence (AI) models often obscure how this noise impacts model accuracy, complicating the optimization of privacy–accuracy trade-offs. This study introduces the Differential Privacy-Hyperdimensional Computing (DP-HD) framework, a novel approach combining Explainable AI (XAI) and vector symbolic paradigms to quantify and predict noise effects on accuracy using a Signal-to-Noise Ratio (SNR) metric. DP-HD enables precise tuning of DP noise levels, ensuring an optimal balance between privacy and performance. The framework has been validated using real-world AM data, demonstrating its applicability to industrial environments. Experimental results demonstrate DP-HD’s capability to achieve state-of-the-art accuracy (94.43%) with robust privacy protections in anomaly detection for AM, even under significant noise conditions. Beyond AM, DP-HD holds substantial promise for broader applications in privacy-sensitive domains such as healthcare, financial services, and government data management, where securing sensitive data while maintaining high ML performance is paramount.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"147 ","pages":"Article 110282"},"PeriodicalIF":7.5,"publicationDate":"2025-02-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143471426","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Meta-machine learning framework for robust short-term solar power prediction across different climatic zones

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-22 DOI: 10.1016/j.engappai.2025.110295

Amit Rai , Ashish Shrivastava , Kartick C. Jana , Jay Liu , Kulwant Singh , N.S. Jayalakshmi , Amit Agrawal

The global energy landscape is increasingly dominated by solar power installations, driven by the sun’s position as Earth’s most abundant and sustainable energy resource. However, the intermittent nature of solar radiation, influenced by both astronomical cycles and meteorological conditions, creates significant challenges for reliable power generation and grid integration. To address the issue of uncertainty, this study proposes a robust and improved capacity machine learning framework with enhanced hypothesis functional space. The proposed model improves the capacity of an individual model by combining the hypothesis functions of individual machine learning models, increasing the representational capacity and hence the model’s generalization. Moreover, a non-linear second stage is stacked to increase the depth of the proposed model, which utilizes meta-data of first stage to further improve the forecasting accuracy. Furthermore, the proposed model is validated on four different climatic zones of the world for solar power forecasting. The proposed model achieves an average improvement of 66.7% in mean absolute error across all locations compared to the next best performing algorithm, with particularly strong performance in arid zones. Statistical validation through Cook’s distance analysis also confirms the model’s reliability with an average of 8.64% influential points across all locations.

{"title":"Meta-machine learning framework for robust short-term solar power prediction across different climatic zones","authors":"Amit Rai , Ashish Shrivastava , Kartick C. Jana , Jay Liu , Kulwant Singh , N.S. Jayalakshmi , Amit Agrawal","doi":"10.1016/j.engappai.2025.110295","DOIUrl":"10.1016/j.engappai.2025.110295","url":null,"abstract":"<div><div>The global energy landscape is increasingly dominated by solar power installations, driven by the sun’s position as Earth’s most abundant and sustainable energy resource. However, the intermittent nature of solar radiation, influenced by both astronomical cycles and meteorological conditions, creates significant challenges for reliable power generation and grid integration. To address the issue of uncertainty, this study proposes a robust and improved capacity machine learning framework with enhanced hypothesis functional space. The proposed model improves the capacity of an individual model by combining the hypothesis functions of individual machine learning models, increasing the representational capacity and hence the model’s generalization. Moreover, a non-linear second stage is stacked to increase the depth of the proposed model, which utilizes meta-data of first stage to further improve the forecasting accuracy. Furthermore, the proposed model is validated on four different climatic zones of the world for solar power forecasting. The proposed model achieves an average improvement of 66.7% in mean absolute error across all locations compared to the next best performing algorithm, with particularly strong performance in arid zones. Statistical validation through Cook’s distance analysis also confirms the model’s reliability with an average of 8.64% influential points across all locations.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"147 ","pages":"Article 110295"},"PeriodicalIF":7.5,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143471368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

One-dimensional input space modelling of a simplified general type-2 Mamdani and Takagi–Sugeno Fuzzy Proportional Integral Derivative controller

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-22 DOI: 10.1016/j.engappai.2025.110289

Ritu Raj

A wide variety of fuzzy controllers have evolved over several decades. Several mathematical models of type-1 and interval type-2 fuzzy controllers have been explored. Most of these modelling approaches involved two-/three- dimensional input space. In this work, we have presented a simplified modelling approach for a General Type-2 (GT2) Mamdani and Takagi–Sugeno (TS) fuzzy Proportional Integral Derivative (PID) controllers involving input space of one-dimension. The fuzzy PID controller’s structure is the parallel combination of the Fuzzy Proportional (FP) plus the Fuzzy Integral (FI) plus the Fuzzy Derivative (FD) control actions. Having a parallel PID control structure simplifies the fuzzy ‘

I F - T H E N

’ rules by eliminating the role of ‘

A N D

’ (triangular norms) and ‘

O R

’ (triangular co-norms) operators. This decoupled rule base aids in decreasing the computing complexity of the GT2 fuzzy PID controller. Owing to the one-dimensional input space, the number of tuneable parameters for fuzzy controllers reduces significantly when compared to two- or three-dimensional input spaces. It is also demonstrated that the type-1 (T1) and interval type-2 (IT2) fuzzy controllers are variations of the GT2 fuzzy controller. In order to assess the controller models, we simulate two systems: the unstable first-order system with dead time and the Continuously Stirred Tank Reactor (CSTR). These models, nevertheless, may also be applied to other dynamic processes and systems.

几十年来，各种模糊控制器不断发展。人们对第一类和第二类区间模糊控制器的若干数学模型进行了探索。这些建模方法大多涉及二维/三维输入空间。在这项工作中，我们提出了一种简化的建模方法，用于涉及一维输入空间的通用第二类（GT2）Mamdani 和高木-菅野（TS）模糊比例积分微分（PID）控制器。模糊 PID 控制器的结构是模糊比例（FP）加模糊积分（FI）加模糊微分（FD）控制动作的并行组合。通过消除 "AND"（三角准则）和 "OR"（三角共准则）运算符的作用，并行 PID 控制结构简化了模糊 "IF-THEN "规则。这种解耦规则库有助于降低 GT2 模糊 PID 控制器的计算复杂度。由于采用了一维输入空间，与二维或三维输入空间相比，模糊控制器的可调参数数量大大减少。研究还证明，类型-1（T1）和区间类型-2（IT2）模糊控制器是 GT2 模糊控制器的变体。为了评估控制器模型，我们模拟了两个系统：具有死区时间的不稳定一阶系统和连续搅拌槽反应器（CSTR）。不过，这些模型也可用于其他动态过程和系统。

{"title":"One-dimensional input space modelling of a simplified general type-2 Mamdani and Takagi–Sugeno Fuzzy Proportional Integral Derivative controller","authors":"Ritu Raj","doi":"10.1016/j.engappai.2025.110289","DOIUrl":"10.1016/j.engappai.2025.110289","url":null,"abstract":"<div><div>A wide variety of fuzzy controllers have evolved over several decades. Several mathematical models of type-1 and interval type-2 fuzzy controllers have been explored. Most of these modelling approaches involved two-/three- dimensional input space. In this work, we have presented a simplified modelling approach for a General Type-2 (GT2) Mamdani and Takagi–Sugeno (TS) fuzzy Proportional Integral Derivative (PID) controllers involving input space of one-dimension. The fuzzy PID controller’s structure is the parallel combination of the Fuzzy Proportional (FP) plus the Fuzzy Integral (FI) plus the Fuzzy Derivative (FD) control actions. Having a parallel PID control structure simplifies the fuzzy ‘<span><math><mrow><mi>I</mi><mi>F</mi><mo>−</mo><mi>T</mi><mi>H</mi><mi>E</mi><mi>N</mi></mrow></math></span>’ rules by eliminating the role of ‘<span><math><mrow><mi>A</mi><mi>N</mi><mi>D</mi></mrow></math></span>’ (triangular norms) and ‘<span><math><mrow><mi>O</mi><mi>R</mi></mrow></math></span>’ (triangular co-norms) operators. This decoupled rule base aids in decreasing the computing complexity of the GT2 fuzzy PID controller. Owing to the one-dimensional input space, the number of tuneable parameters for fuzzy controllers reduces significantly when compared to two- or three-dimensional input spaces. It is also demonstrated that the type-1 (T1) and interval type-2 (IT2) fuzzy controllers are variations of the GT2 fuzzy controller. In order to assess the controller models, we simulate two systems: the unstable first-order system with dead time and the Continuously Stirred Tank Reactor (CSTR). These models, nevertheless, may also be applied to other dynamic processes and systems.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"147 ","pages":"Article 110289"},"PeriodicalIF":7.5,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143465531","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Hybrid deep-learning prediction model based on kernel multi-granularity fuzzy rough sets and its application in the diagnosis and treatment of chronic kidney disease

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-22 DOI: 10.1016/j.engappai.2025.110297

Jiqian Liu , Bingzhen Sun , Jin Ye , Xixuan Zhao , Xiaoli Chu

In uncertain decision-making scenarios, quantitative scientific prediction models and methods can provide valuable support for making scientific decisions. However, the characteristics of hybrid attribute information may lead to a series of issues. These include difficulties in comparing and comprehensively evaluating different types of attributes, nonlinear relationships between attributes, and a lack of effective decision-support methods. To overcome these issues, this study introduces a kernel function to abstract the similarity of different attribute types and proposes a model called kernel multi-granularity fuzzy rough sets (KMGFRS). The KMGFRS model facilitates a thorough exploration and analysis of the uncertainties associated with decision objects. Additionally, an attribute reduction method based on KMGFRS is discussed to address redundant attributes in hybrid information systems. This method eliminates attributes that have a minimal influence on the decision results, simplifies the decision process, and enhances its effectiveness. This study integrates the KMGFRS and hybrid deep learning concepts to propose a novel prediction method aimed at enhancing accuracy and robustness. From the perspective of hybrid attribute information, this method can more accurately predict the unknown attributes of decision objects, thereby providing robust support for disease prediction in medical diagnostics and therapeutic decision-making. The experimental results indicated that the constructed model effectively handled uncertain decision-making scenarios involving hybrid attributes and fuzzy decision objects. It provides accurate and reliable decision support for chronic kidney disease (CKD), significantly enhancing the predictive accuracy of CKD types.

{"title":"Hybrid deep-learning prediction model based on kernel multi-granularity fuzzy rough sets and its application in the diagnosis and treatment of chronic kidney disease","authors":"Jiqian Liu , Bingzhen Sun , Jin Ye , Xixuan Zhao , Xiaoli Chu","doi":"10.1016/j.engappai.2025.110297","DOIUrl":"10.1016/j.engappai.2025.110297","url":null,"abstract":"<div><div>In uncertain decision-making scenarios, quantitative scientific prediction models and methods can provide valuable support for making scientific decisions. However, the characteristics of hybrid attribute information may lead to a series of issues. These include difficulties in comparing and comprehensively evaluating different types of attributes, nonlinear relationships between attributes, and a lack of effective decision-support methods. To overcome these issues, this study introduces a kernel function to abstract the similarity of different attribute types and proposes a model called kernel multi-granularity fuzzy rough sets (KMGFRS). The KMGFRS model facilitates a thorough exploration and analysis of the uncertainties associated with decision objects. Additionally, an attribute reduction method based on KMGFRS is discussed to address redundant attributes in hybrid information systems. This method eliminates attributes that have a minimal influence on the decision results, simplifies the decision process, and enhances its effectiveness. This study integrates the KMGFRS and hybrid deep learning concepts to propose a novel prediction method aimed at enhancing accuracy and robustness. From the perspective of hybrid attribute information, this method can more accurately predict the unknown attributes of decision objects, thereby providing robust support for disease prediction in medical diagnostics and therapeutic decision-making. The experimental results indicated that the constructed model effectively handled uncertain decision-making scenarios involving hybrid attributes and fuzzy decision objects. It provides accurate and reliable decision support for chronic kidney disease (CKD), significantly enhancing the predictive accuracy of CKD types.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"147 ","pages":"Article 110297"},"PeriodicalIF":7.5,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143471369","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Drug–target affinity prediction using rotary encoding and information retention mechanisms 利用旋转编码和信息保留机制预测药物与靶点的亲和力

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-22 DOI: 10.1016/j.engappai.2025.110239

Zhiqin Zhu , Yan Ding , Guanqiu Qi , Baisen Cong , Yuanyuan Li , Litao Bai , Xinbo Gao

Drug–target affinity (DTA) prediction has been widely used in pharmaceutical research as a novel and effective method to explore the interaction strength between drugs and targets. However, existing DTA prediction models mainly rely on graphical representations of drug molecules, overlooking the intricate interactions between individual substructures. This limitation impacts both the predictive accuracy and the informational richness within the model nodes. To address these challenges, this paper proposes the Rotary Retention Graph Drug–Target Affinity (RRGDTA) network with rotation and retention mechanisms. The RRGDTA integrates an information interaction module into the extraction of drug and target features across multiple scale levels. This approach enhances the correlation within the graph representation, leading to an optimal feature representation. Furthermore, to tackle the issue of limited relationship between molecular structure and context, a Multi-Scale Interaction module (MSI) is proposed to enhance important features related to both. Additionally, to address inaccuracies in the structural features of drugs and targets, a Rotary Encoding Module (ROE) is proposed, which focuses on nearby contextual information and effectively captures the correlation between them. In order to solve the problem of insufficient information representation, an Association Prediction Module (APM) and an Intra-Mask Retention Module (IMR) are proposed to maximize the retention of drug–target information. The efficacy of the proposed RRGDTA in DTA prediction was validated on the Davis, kinase inhibitors biochemical assays (KIBA) and binding database (BindingDB) datasets. Compared with current baseline models, the proposed model achieved better results across various metrics, demonstrating its superior performance in accurate DTA prediction.

{"title":"Drug–target affinity prediction using rotary encoding and information retention mechanisms","authors":"Zhiqin Zhu , Yan Ding , Guanqiu Qi , Baisen Cong , Yuanyuan Li , Litao Bai , Xinbo Gao","doi":"10.1016/j.engappai.2025.110239","DOIUrl":"10.1016/j.engappai.2025.110239","url":null,"abstract":"<div><div>Drug–target affinity (DTA) prediction has been widely used in pharmaceutical research as a novel and effective method to explore the interaction strength between drugs and targets. However, existing DTA prediction models mainly rely on graphical representations of drug molecules, overlooking the intricate interactions between individual substructures. This limitation impacts both the predictive accuracy and the informational richness within the model nodes. To address these challenges, this paper proposes the Rotary Retention Graph Drug–Target Affinity (RRGDTA) network with rotation and retention mechanisms. The RRGDTA integrates an information interaction module into the extraction of drug and target features across multiple scale levels. This approach enhances the correlation within the graph representation, leading to an optimal feature representation. Furthermore, to tackle the issue of limited relationship between molecular structure and context, a Multi-Scale Interaction module (MSI) is proposed to enhance important features related to both. Additionally, to address inaccuracies in the structural features of drugs and targets, a Rotary Encoding Module (ROE) is proposed, which focuses on nearby contextual information and effectively captures the correlation between them. In order to solve the problem of insufficient information representation, an Association Prediction Module (APM) and an Intra-Mask Retention Module (IMR) are proposed to maximize the retention of drug–target information. The efficacy of the proposed RRGDTA in DTA prediction was validated on the Davis, kinase inhibitors biochemical assays (KIBA) and binding database (BindingDB) datasets. Compared with current baseline models, the proposed model achieved better results across various metrics, demonstrating its superior performance in accurate DTA prediction.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"147 ","pages":"Article 110239"},"PeriodicalIF":7.5,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143471371","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Few-shot machine reading comprehension for bridge inspection via domain-specific and task-aware pre-tuning approach

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-22 DOI: 10.1016/j.engappai.2025.110361

Ren Li , Luyi Zhang , Qiao Xiao , Jianxi Yang , Yu Chen , Shixin Jiang , Di Wang

With the wide application of information technologies in the field of bridge engineering, many electronic bridge inspection reports have been generated. However, due to insufficient research on machine reading comprehension (MRC) in this field, a lot of bridge inspection information, e.g., structural basic data, inspected defects, and maintenance suggestions, has not been fully used. Especially, it is time-consuming and labor-intensive to pre-train a domain-specific language model from scratch or annotate large-scale question answering corpora, which also brings challenges to the MRC research in this field. To tackle the problems, this paper proposes a novel few-shot MRC approach for bridge inspection based on the idea of data augmentation. The proposed model uses a pre-trained model as backbone, along with introducing a pre-tuning stage to bridge the gaps between general-purpose pre-training and domain-specific MRC tasks. In order to reduce the workload of manual annotation, we present a novel pre-tuning data generation algorithm which is based on the domain-specific question classification and answer prediction neural models. After pre-tuning and fine-tuning, the proposed model achieves efficient bridge inspection MRC. The experimental results show that the proposed model outperforms the mainstream fine-tuning-based approaches and few-shot MRC baseline models in various settings. With 1024 fine-tuning samples, the F1 value and Exact Match (EM) value are 86.42%, 74.65%, respectively. Our research work can serve as a foundation for the construction of automatic question answering systems for intelligent bridge management and maintenance.

{"title":"Few-shot machine reading comprehension for bridge inspection via domain-specific and task-aware pre-tuning approach","authors":"Ren Li , Luyi Zhang , Qiao Xiao , Jianxi Yang , Yu Chen , Shixin Jiang , Di Wang","doi":"10.1016/j.engappai.2025.110361","DOIUrl":"10.1016/j.engappai.2025.110361","url":null,"abstract":"<div><div>With the wide application of information technologies in the field of bridge engineering, many electronic bridge inspection reports have been generated. However, due to insufficient research on machine reading comprehension (MRC) in this field, a lot of bridge inspection information, e.g., structural basic data, inspected defects, and maintenance suggestions, has not been fully used. Especially, it is time-consuming and labor-intensive to pre-train a domain-specific language model from scratch or annotate large-scale question answering corpora, which also brings challenges to the MRC research in this field. To tackle the problems, this paper proposes a novel few-shot MRC approach for bridge inspection based on the idea of data augmentation. The proposed model uses a pre-trained model as backbone, along with introducing a pre-tuning stage to bridge the gaps between general-purpose pre-training and domain-specific MRC tasks. In order to reduce the workload of manual annotation, we present a novel pre-tuning data generation algorithm which is based on the domain-specific question classification and answer prediction neural models. After pre-tuning and fine-tuning, the proposed model achieves efficient bridge inspection MRC. The experimental results show that the proposed model outperforms the mainstream fine-tuning-based approaches and few-shot MRC baseline models in various settings. With 1024 fine-tuning samples, the F1 value and Exact Match (EM) value are 86.42%, 74.65%, respectively. Our research work can serve as a foundation for the construction of automatic question answering systems for intelligent bridge management and maintenance.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"147 ","pages":"Article 110361"},"PeriodicalIF":7.5,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143465110","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

On-board detection of rail corrugation using improved convolutional block attention mechanism 利用改进的卷积块注意机制对轨道波纹进行车载检测

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-22 DOI: 10.1016/j.engappai.2025.110349

Yang Wang , Hong Xiao , Chaozhi Ma , Zhihai Zhang , Xuhao Cui , Aimin Xu

Leveraging acceleration sensors affixed to the train body enables continuous surveillance of rail corrugation, delivering cost-effectiveness, operational efficiency, and portability. Establishing the correlation between vertical body acceleration and rail corrugation poses a substantial challenge. To ensure uninterrupted monitoring of rail corrugation, an initial development involved constructing a train-track integrated simulation model that accounted for the dynamics of flexible wheelsets and tracks, thereby generating a simulated dataset of vertical body acceleration. Subsequent improvements were made to the conventional Convolutional Block Attention Module (CBAM) architecture, culminating in the proposal of a deep one-dimensional convolutional residual network model named Train Body Vertical Acceleration Network (TBVA-Net), founded on an improved CBAM framework. Training was conducted using the simulated dataset, showcasing the reduced model complexity and total parameter count of the improved CBAM architecture, which notably amplified classification accuracy. The TBVA-Net, employing the refined CBAM, consistently achieved test accuracies exceeding 95%, averaging at 98.6% on the simulated dataset. Validation through field-measured data corroborated the rationale behind the proposed TBVA-Net architecture. Fine-tuning with a limited subset of labeled field data led to a transfer accuracy of 98.5%. This paper presents an innovative approach for detecting rail corrugation through vertical acceleration signals obtained from operational vehicles.

{"title":"On-board detection of rail corrugation using improved convolutional block attention mechanism","authors":"Yang Wang , Hong Xiao , Chaozhi Ma , Zhihai Zhang , Xuhao Cui , Aimin Xu","doi":"10.1016/j.engappai.2025.110349","DOIUrl":"10.1016/j.engappai.2025.110349","url":null,"abstract":"<div><div>Leveraging acceleration sensors affixed to the train body enables continuous surveillance of rail corrugation, delivering cost-effectiveness, operational efficiency, and portability. Establishing the correlation between vertical body acceleration and rail corrugation poses a substantial challenge. To ensure uninterrupted monitoring of rail corrugation, an initial development involved constructing a train-track integrated simulation model that accounted for the dynamics of flexible wheelsets and tracks, thereby generating a simulated dataset of vertical body acceleration. Subsequent improvements were made to the conventional Convolutional Block Attention Module (CBAM) architecture, culminating in the proposal of a deep one-dimensional convolutional residual network model named Train Body Vertical Acceleration Network (TBVA-Net), founded on an improved CBAM framework. Training was conducted using the simulated dataset, showcasing the reduced model complexity and total parameter count of the improved CBAM architecture, which notably amplified classification accuracy. The TBVA-Net, employing the refined CBAM, consistently achieved test accuracies exceeding 95%, averaging at 98.6% on the simulated dataset. Validation through field-measured data corroborated the rationale behind the proposed TBVA-Net architecture. Fine-tuning with a limited subset of labeled field data led to a transfer accuracy of 98.5%. This paper presents an innovative approach for detecting rail corrugation through vertical acceleration signals obtained from operational vehicles.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"146 ","pages":"Article 110349"},"PeriodicalIF":7.5,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143464861","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SPARDA: Sparsity-constrained dimensional analysis via convex relaxation for parameter reduction in high-dimensional engineering systems

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-22 DOI: 10.1016/j.engappai.2025.110307

Kuang Yang, Qiang Li, Zhenghui Hou, Haifan Liao, Chaofan Yang, Haijun Wang

Effective analysis of high-dimensional systems with intricate variable interactions is crucial for accurate modeling and engineering applications. Previous methods using sparsity techniques or dimensional analysis separately often face limitations when handling complex, large-scale systems. This study introduces a sparsity-constrained dimensional analysis framework that integrates the classical Buckingham Pi theorem with sparse optimization techniques, enabling precise nondimensionalization. The framework, formulated as a convex optimization problem, addresses computational challenges associated with sparsity in high-dimensional spaces. Rigorously tested across various datasets, including the Fanning friction factor for rough pipe flow, an international standards-based dataset of physical quantities and units, and experimental data from flow boiling studies, this method successfully identified critical dimensionless groups that encapsulate core system dynamics. This approach not only offers a more compact and interpretable representation than conventional methods but also retains more characteristics of function variability. It proves particularly effective in systems governed by high-dimensional interactions, demonstrating a lower failure rate and mean relative error compared to an algorithm for comparison. The methodology is applicable to the modeling and analysis of complex engineering physical systems such as nuclear power, wind tunnel design, and marine engineering, as well as in designing scaled verification experiments.

{"title":"SPARDA: Sparsity-constrained dimensional analysis via convex relaxation for parameter reduction in high-dimensional engineering systems","authors":"Kuang Yang, Qiang Li, Zhenghui Hou, Haifan Liao, Chaofan Yang, Haijun Wang","doi":"10.1016/j.engappai.2025.110307","DOIUrl":"10.1016/j.engappai.2025.110307","url":null,"abstract":"<div><div>Effective analysis of high-dimensional systems with intricate variable interactions is crucial for accurate modeling and engineering applications. Previous methods using sparsity techniques or dimensional analysis separately often face limitations when handling complex, large-scale systems. This study introduces a sparsity-constrained dimensional analysis framework that integrates the classical Buckingham Pi theorem with sparse optimization techniques, enabling precise nondimensionalization. The framework, formulated as a convex optimization problem, addresses computational challenges associated with sparsity in high-dimensional spaces. Rigorously tested across various datasets, including the Fanning friction factor for rough pipe flow, an international standards-based dataset of physical quantities and units, and experimental data from flow boiling studies, this method successfully identified critical dimensionless groups that encapsulate core system dynamics. This approach not only offers a more compact and interpretable representation than conventional methods but also retains more characteristics of function variability. It proves particularly effective in systems governed by high-dimensional interactions, demonstrating a lower failure rate and mean relative error compared to an algorithm for comparison. The methodology is applicable to the modeling and analysis of complex engineering physical systems such as nuclear power, wind tunnel design, and marine engineering, as well as in designing scaled verification experiments.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"146 ","pages":"Article 110307"},"PeriodicalIF":7.5,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143464860","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A novel enhanced Superlet Synchroextracting transform ensemble learning for structural health monitoring using nonlinear wave modulation

IF 7.5 2区计算机科学 Q1 AUTOMATION & CONTROL SYSTEMS

Engineering Applications of Artificial Intelligence

Pub Date : 2025-02-22 DOI: 10.1016/j.engappai.2025.110341

Naserodin Sepehry , Mohammad Ehsani , Hamdireza Amindavar , Weidong Zhu , Firooz Bakhtiari Nejad

This study investigates the application of nonlinear wave modulation (NWM) using chirp signals for structural health monitoring (SHM). The implementation of NWM with monoharmonic signals (periodic signals that consist of a single frequency component) poses significant challenges due to the complexity of selecting optimal pump and carrier frequencies, leading to time-intensive processes. In contrast, analyzing NWM with chirp signals introduces additional complexities regarding signal processing compared to monoharmonic excitations. Time-frequency analysis (TFA) has been identified as a crucial method for examining non-stationary signals; however, many existing techniques face limitations in resolution, particularly in the context of chirp signals, as dictated by the Heisenberg uncertainty principle. To address these challenges, the superlet synchroextracting transform (SLSET) is introduced as an innovative TFA approach that combines the strengths of superlet (SL) and synchroextracting transforms, resulting in improved resolution. This research utilizes NWM alongside SLSET to detect boundary loosening in sandwich beams, demonstrating the method's effectiveness in identifying structural damage while maintaining robustness against noise. Results indicate that SLSET significantly enhances the damage index compared to traditional TFA methods. The high resolution achieved allows for the detection of sidebands in vibro-acoustic modulation (VAM) tests conducted at low pump frequencies. Furthermore, three machine learning (ML) models including support vector machine (SVM), Adaptive Boosting (AdaBoost), and Random Forest (RF) were trained. The stack ensemble method combined the outputs of these models, resulting in an overall accuracy of 99.2%. This approach effectively leveraged the strengths of individual models, enhancing generalization and robustness in detecting damage across complex data scenarios. The features extracted using SLSET for VAM data of faulty structure attains a classification accuracy of 98.9%. In contrast, features derived from conventional time-frequency methods fail to identify damage, even in noise-free conditions.

{"title":"A novel enhanced Superlet Synchroextracting transform ensemble learning for structural health monitoring using nonlinear wave modulation","authors":"Naserodin Sepehry , Mohammad Ehsani , Hamdireza Amindavar , Weidong Zhu , Firooz Bakhtiari Nejad","doi":"10.1016/j.engappai.2025.110341","DOIUrl":"10.1016/j.engappai.2025.110341","url":null,"abstract":"<div><div>This study investigates the application of nonlinear wave modulation (NWM) using chirp signals for structural health monitoring (SHM). The implementation of NWM with monoharmonic signals (periodic signals that consist of a single frequency component) poses significant challenges due to the complexity of selecting optimal pump and carrier frequencies, leading to time-intensive processes. In contrast, analyzing NWM with chirp signals introduces additional complexities regarding signal processing compared to monoharmonic excitations. Time-frequency analysis (TFA) has been identified as a crucial method for examining non-stationary signals; however, many existing techniques face limitations in resolution, particularly in the context of chirp signals, as dictated by the Heisenberg uncertainty principle. To address these challenges, the superlet synchroextracting transform (SLSET) is introduced as an innovative TFA approach that combines the strengths of superlet (SL) and synchroextracting transforms, resulting in improved resolution. This research utilizes NWM alongside SLSET to detect boundary loosening in sandwich beams, demonstrating the method's effectiveness in identifying structural damage while maintaining robustness against noise. Results indicate that SLSET significantly enhances the damage index compared to traditional TFA methods. The high resolution achieved allows for the detection of sidebands in vibro-acoustic modulation (VAM) tests conducted at low pump frequencies. Furthermore, three machine learning (ML) models including support vector machine (SVM), Adaptive Boosting (AdaBoost), and Random Forest (RF) were trained. The stack ensemble method combined the outputs of these models, resulting in an overall accuracy of 99.2%. This approach effectively leveraged the strengths of individual models, enhancing generalization and robustness in detecting damage across complex data scenarios. The features extracted using SLSET for VAM data of faulty structure attains a classification accuracy of 98.9%. In contrast, features derived from conventional time-frequency methods fail to identify damage, even in noise-free conditions.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"147 ","pages":"Article 110341"},"PeriodicalIF":7.5,"publicationDate":"2025-02-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143465111","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0