Article
Automation & Control Systems
M. Mowbray, H. Kay, S. Kay, P. Castro Caetano, A. Hicks, C. Mendoza, A. Lane, P. Martin, D. Zhang
Summary: This study combines latent variable modeling with probabilistic machine learning methods to develop novel soft-sensors for industrial batch process monitoring. Experimental validation shows that the MPLS-HNN soft-sensor exhibits high accuracy, reliability, and practical implementation.
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
(2022)
Review
Automation & Control Systems
Yasith S. Perera, D. A. A. C. Ratnaweera, Chamila H. Dasanayaka, Chamil Abeykoon
Summary: With the depletion of natural resources and environmental issues, the concept of sustainable development has become important in process industries. Manufacturers are adopting novel process monitoring techniques, including soft sensors, to enhance product quality and efficiency while minimizing adverse environmental impacts. This article explores the role of AI-driven soft sensors in achieving sustainable development goals in process industries.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Minjung Lee, Jinsoo Bae, Seoung Bum Kim
Summary: Data-driven soft sensors using deep learning models have shown superior predictive performance, but may face trustworthiness issues when dealing with unexpected situations or noisy input data. By introducing uncertainty-aware soft sensors based on Bayesian recurrent neural networks, the reliability of predictive uncertainty can be increased, allowing for interval prediction without compromising the predictive performance of the soft sensor.
ADVANCED ENGINEERING INFORMATICS
(2021)
Article
Chemistry, Analytical
Xiankai Cheng, Benkun Bao, Weidong Cui, Shuai Liu, Jun Zhong, Liming Cai, Hongbo Yang
Summary: In this study, we used a wireless miniaturized inertial navigation sensors (WMINS) network to obtain limb movement information in a virtual environment. By processing the data features and analyzing motion characteristics, we developed a classification model for acrophobia and non-acrophobia and successfully recognized and classified them using the designed integrated learning model. The final accuracy of dichotomous classification based on limb motion information reached 94.64%, showing higher accuracy and efficiency compared to existing research models. Overall, our study demonstrates a strong correlation between people's mental state during fear of heights and their limb movements at that time.
Article
Chemistry, Multidisciplinary
Yang Luo, Xiao Xiao, Jun Chen, Qian Li, Hongyan Fu
Summary: Soft interfaces with self-sensing capabilities are crucial for environment awareness and reaction. This study presents a bioinspired soft sensor array (BOSSA) that integrates pressure and material sensing based on the triboelectric effect. BOSSA demonstrates high accuracy in user identification and object placement/extraction, and its scalable fabrication enables large-area sensor arrays with high resolution and multimodal sensing abilities. This research has potential implications for intelligent monitoring and stimuli response in various applications.
Article
Computer Science, Artificial Intelligence
Naeem Jan, Jeonghwan Gwak, Dragan Pamucar
Summary: Generative Adversarial Networks (GANs) are models that generate data samples based on the statistical distribution of the data, and they belong to one of the most well-known branches in machine learning and deep learning. This paper introduces a solid mathematical concept to model and resolve the problem of complex picture fuzzy soft relations (CPFSRs) by combining two different theories, picture fuzzy set (PFS) and soft set (SS). The proposed methodology, based on CPFSRs, is used for analyzing and selecting the best GAN for effective working, and a comparative study of existing techniques has been conducted to validate the proposed work.
APPLIED SOFT COMPUTING
(2023)
Article
Engineering, Multidisciplinary
Yaoyao Bao, Yuanming Zhu, Feng Qian
Summary: Inspired by meta-learning achievements, this paper proposes a local quadratic embedding learning algorithm for regression problems using metric learning and neural networks. The algorithm optimizes Mahalanobis metric learning and introduces lightweight NNs to learn coefficient matrices and assign weights. The algorithm prevents model degradation caused by sensor drift and unmeasurable variables, and outperforms popular regression methods.
Article
Chemistry, Analytical
Mounir Guesbaya, Francisco Garcia-Manas, Francisco Rodriguez, Hassina Megherbi
Summary: This paper presents the development of a soft sensor based on LSTM-RNN to estimate the opening of vents in traditional Mediterranean greenhouses. The results demonstrate that the developed soft sensor can accurately estimate the actual opening of the vents with a mean absolute error of 4.45%.
Article
Automation & Control Systems
Wu-Te Yang, Masayoshi Tomizuka
Summary: This paper introduces a multifunctional soft tactile sensor that can estimate contact force, contact feature, and contact point simultaneously. The sensor has a dual-layer structure with the top layer detecting contact location and feature, and the bottom layer measuring contact force. Experimental results verify its effectiveness in capturing contact information of fragile objects.
JOURNAL OF DYNAMIC SYSTEMS MEASUREMENT AND CONTROL-TRANSACTIONS OF THE ASME
(2022)
Article
Engineering, Industrial
Y. Mizutani, S. Kataoka, Y. Nagai, T. Uenohara, Y. Takaya
Summary: The structure of deep neural networks (DNNs) used in triangulation displacement sensors was investigated through numerical and experimental analysis. A numerical model was constructed by adding noise to the ideal waveform of the measurements, and the major components of the DNNs were optimized through numerical calculations. The optimized DNNs were then applied in a measurement system for distance measurement with sub-pixel accuracy.
CIRP ANNALS-MANUFACTURING TECHNOLOGY
(2022)
Article
Physics, Multidisciplinary
Lijun Li, Mengge Xue, Tianzong Xu, Yinming Liu, Yibo Yuan, Zheng Lin
Summary: This paper introduces an all-fiber flexible soft sensor that combines machine learning methods to recognize its own twisted shape. With high accuracy, simple structure, and reliable performance, the sensor shows great potential in intelligent robots, medical rehabilitation, surgical endoscopes, and object recognition.
Article
Engineering, Multidisciplinary
Pasquale Arpaia, Renato Cuocolo, Francesco Donnarumma, Antonio Esposito, Nicola Moccaldi, Angela Natalizio, Roberto Prevete
Summary: With the increasing number of elderly people, heart diseases have become a major issue in healthy aging. A concept design of a soft sensor for measuring cardiovascular risk in real time showed promising results with high classification accuracy in the experiment.
Review
Engineering, Environmental
Phoebe M. L. Ching, Richard H. Y. So, Tobias Morck
Summary: The use of soft sensors in wastewater treatment plants has advanced significantly, from mechanistic modelling to machine learning models. Neural networks have been the dominant methodology for soft sensor development, but decision tree-based approaches have shown promising performance and enhanced robustness. Utilizing soft sensor modelling approaches can enhance hardware sensor performance, leading to continuous improvements in reliability and measurement range.
JOURNAL OF WATER PROCESS ENGINEERING
(2021)
Article
Materials Science, Multidisciplinary
Jun-Hyoung Park, Ji-Ho Cho, Jung-Sik Yoon, Jung-Ho Song
Summary: The non-invasive approach presented in this study utilizes optical emission spectroscopy (OES) and multivariate data analysis to monitor plasma parameters inside a radio-frequency (RF) plasma nitridation device. An empirical correlation was established for real-time monitoring using machine learning (ML) based on simultaneous OES and other diagnostics, achieving high prediction accuracy for electron density and temperature. This method provides in-situ and real-time analysis without disturbing the plasma or interfering with the process, making it especially useful in plasma processing.
Article
Automation & Control Systems
Ze Yang Ding, Junn Yong Loo, Surya Girinatha Nurzaman, Chee Pin Tan, Vishnu Monn Baskaran
Summary: This article proposes a deep learning-based modeling framework for developing soft sensor models that are robust to sensor faults. The framework uses a zero-shot learning approach, training the model with only fault-free dataset to save time and resources. Adversarial examples are used for modeling faulty sensor inputs, enabling the model to be adaptive and achieve robustness.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Marcela Vallejo, Carlos J. Gallego, L. Duque-Munoz, Edilson Delgado-Trejos
Article
Engineering, Electrical & Electronic
Luis Enrique Avendano, Luis David Avendano-Valencia, Edilson Delgado-Trejos
Review
Genetics & Heredity
Manuel Mauricio Goez, Maria Constanza Torres-Madronero, Sarah Rothlisberger, Edilson Delgado-Trejos
GENOMICS PROTEOMICS & BIOINFORMATICS
(2018)
Article
Physics, Multidisciplinary
David Cuesta-Frau, Juan Pablo Murillo-Escobar, Diana Alexandra Orrego, Edilson Delgado-Trejos
Article
Engineering, Electrical & Electronic
Luis Enrique Avendano, Luis David Avendano-Valencia, Edilson Delgado-Trejos, David Cuesta-Frau
Summary: This work focuses on the harmonic decomposition of pseudo-periodic non-stationary multivariate signals. The proposed method achieves this by establishing a procedure to transform the multivariate signal into a block-diagonal state-space representation. The harmonic components and instantaneous frequency are estimated using Kalman filtering, and an optimization framework is provided for the hyperparameters of the state space representation.
Article
Computer Science, Information Systems
Francisco Villa, Cherlly Sanchez, Marcela Vallejo, Juan S. Botero-Valencia, Edilson Delgado-Trejos
Summary: The analysis of flow-induced pipe vibrations has been applied in various applications, such as flowrate inference and leak detection. This study collected a dataset of signals from an accelerometer attached to a pipe conveying cold water, with a total of 382 signals containing acceleration values in three axes and a timestamp in microseconds.
Article
Computer Science, Information Systems
Juan E. Urrea, Luisa F. Restrepo, Jeanette Prada-Arismendy, Erwing Castillo, Manuel M. Goez, Maria C. Torres-Madronero, Edilson Delgado-Trejos, Sarah Rothlisberger
Summary: AML is a malignant disorder of hematopoietic stem and progenitor cells, with accurately assessing patient prognosis being crucial for clinical management. Currently, proteomic changes related to treatment response in AML patients have not been extensively explored.
Proceedings Paper
Computer Science, Artificial Intelligence
C. Duque-Mejia, M. A. Becerra, C. Zapata-Hernandez, C. Mejia-Arboleda, A. E. Castro-Ospina, E. Delgado-Trejos, Diego H. Peluffo-Ordonez, P. Rosero-Montalvo, Javier Revelo-Fuelagan
INTELLIGENT INFORMATION AND DATABASE SYSTEMS, ACIIDS 2019, PT I
(2019)
Article
Computer Science, Information Systems
Carlos A. Duarte-Salazar, Andres Eduardo Castro-Ospina, Miguel A. Becerra, Edilson Delgado-Trejos