Liangze Cui

Mr. Liangze Cui

orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 June 2020 in IEEE Transactions on Antennas and Propagation
Verified
Journal Article
Closed access
LCLiangze Cui
Liangze Cui
YZYao Zhang
Yao Zhang
RZRunren Zhang
Runren Zhang
QLQing Huo Liu
Qing Huo Liu
Effective machine learning methods are usually trained by large datasets to guarantee accuracy and avoid overfitting. However, large datasets restrict the popularization of machine learning methods when the data acquisition is nontrivial. To solve this problem, a novel machine learning method is proposed in this paper based on the modified K-Nearest Neighbors (KNN) algorithm, which can extract more features from datasets through advanced workflow and simulation techniques. In the applications presented here, our method is 5 to 30 times faster than traditional machine learning methods such as artificial neural network (ANN) and Bayesian optimization by reducing the required size of datasets. The proposed method is then employed to optimize the antenna parameters, while an additional branch is built to run the simulation tools (e.g. HFSS) and update the dataset during the training process instead of constructing the dataset beforehand. The validity and efficiency of this proposed method are confirmed by four different antenna examples and other machine learning and gradient-based algorithms. In summary, the proposed method can obtain a satisfactory optimal antenna design at little cost.

11 May 2020 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
RZRunren Zhang
Runren Zhang
QSQingtao Sun
Qingtao Sun
XZXinyuan Zhang
Xinyuan Zhang
LCLiangze Cui
Liangze Cui
ZWZhenguan Wu
Zhenguan Wu
QLQing Huo Liu
Qing Huo Liu
Imaging hydraulic fractures is of paramount importance to subsurface resource extraction, geologic storage, and hazardous waste disposal. The use of electrically conductive proppants and current energized steel casing provides a promising approach to monitor the distribution of fractures. In this article, a borehole-to-surface system is employed to energize the steel casing and measure electric and magnetic fields on the ground. A convolutional neural network (CNN) is then trained to learn the relationship between the measured field pattern and the parameterized fracture, namely, the lateral extent and direction. To accelerate the generation of training data with limited accuracy loss, an approximate hollow casing is modeled by the impedance transition boundary condition with a tenfold magnified radius and reduced conductivity. Two training strategies are then presented with a grid search of the network's hyperparameters. The well-trained CNN shows good generalization to unseen fracture conductivity, the true casing model, as well as white Gaussian noise. Finally, we apply the CNN to image irregular fractures and obtain reliable results even under strong noise, indicating a promising imaging technique for more complicated fractures.

28 March 2022 in IEEE Transactions on Antennas and Propagation
Verified
Journal Article
Closed access
RZRunren Zhang
Runren Zhang
QSQingtao Sun
Qingtao Sun
YMYiqian Mao
Yiqian Mao
LCLiangze Cui
Liangze Cui
YJYongze Jia
Yongze Jia
QLQing Huo Liu
Qing Huo Liu
Deep transfer learning has a great success story in computer vision (CV), natural language processing, and many other fields. In this communication, we are going to push forward the deep transfer learning to the hydraulic fracture imaging problem by proposing a two-step approach: 1) train a convolutional neural network (CNN) to reconstruct target geometries by a relatively large amount of approximated field patterns generated from a simplified model and 2) fine-tune the top layers of transferred CNN by a small amount of true field patterns generated through a full model. The advantages include the rapid generation of large amount of data through the simplified model and the high reconstruction accuracy through a careful design of the deep transfer learning. The CNN trained through the deep transfer learning can accurately reconstruct the lateral extent and direction of fractures with unseen conductivity and white Gaussian noise, showing a notable acceleration/accuracy enhancement over the previous CNN trained by mixing a small number of true data into the approximated data as data augmentation.

1 November 2024 in IEEE Transactions on Antennas and Propagation
Verified
Journal Article
Closed access
LCLiangze Cui
Liangze Cui
KHKai Huang
Kai Huang
YZYang Zhong
Yang Zhong
YZYao Zhang
Yao Zhang
QLQing Huo Liu
Qing Huo Liu
Machine learning (ML) methods are widely employed in antenna design and optimization due to their strong performance in multi-dimensional optimization problems and reduced dependence on engineers’ experience. In this work, we propose an auxiliary classifier generative method inspired by the auxiliary classifier generative adversarial networks (AC-GANs) from the computer vision domain, integrated with an evolutionary strategy to facilitate class-based antenna design. By allocating suitable class labels, multiple design requirements can be efficiently met. This method categorizes antenna designs to enable simultaneous optimization of various antenna types within a unified geometric model. Utilizing a discriminator-generator framework typical of GANs, the discriminator predicts class labels from geometric models, while the generator creates new candidates to satisfy these classifications. An additional filter layer, a random forest classifier, refines the generator’s outputs, proving effective in our experiments. The incorporation of evolutionary criteria significantly enhances the design process, mirroring active learning principles to address the challenges of limited and expensive datasets. The final outcomes include both optimized geometric designs and a robust generator capable of producing these designs in large quantities. We substantiate the efficacy and benefits of our proposed method through an analytical case study and a practical dipole antenna example. The comparison between our method and other optimization methods highlights three distinct advantages of the proposed approach: versatility, scalability, and high efficiency.

20 January 2026 in IEEE Transactions on Antennas and Propagation
Verified
Journal Article
Closed access
LCLiangze Cui
Liangze Cui
KHKai Huang
Kai Huang
YZYang Zhong
Yang Zhong
YZYao Zhang
Yao Zhang
QLQing Huo Liu
Qing Huo Liu
This paper presents a generative machine learning (ML) framework for the automated design and optimization of filtering antennas, a class of antennas that requires the simultaneous control of S-parameters and gain across multiple frequency bands. The proposed method reformulates the inverse design problem as a binary valid/invalid classification task. An auxiliary-classifier generative adversarial network (AC-GAN) is employed to generate candidate designs, while a threshold-moving random forest (TM-RF) classifier serves as a post-processing filter to address the severe data imbalance inherent in filtering-antenna synthesis. An evolutionary refinement strategy further constrains the design space across iterations, improving model stability and convergence. Building upon our earlier work [1] on generative antenna design, this study introduces new mechanisms specifically required for filtering antennas, which exhibit tightly coupled multi-objective constraints and significantly higher simulation costs. The proposed framework is used to design a compact filtering antenna, and the resulting prototype is fabricated and experimentally validated. Measured results demonstrate close agreement with simulations, confirming the effectiveness of the ML-generated design. Physical analysis of current distribution and filtering behavior is also performed to ensure consistency with established antenna theory. Overall, the proposed generative ML framework offers an efficient and scalable approach for the inverse design of complex electromagnetic (EM) structures, particularly filtering antennas with stringent performance requirements.

17 August 2023 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
YZYang Zhong
Yang Zhong
HWHanming Wang
Hanming Wang
WHWei-Feng Huang
Wei-Feng Huang
YMYiqian Mao
Yiqian Mao
MYMengqing Yuan
Mengqing Yuan
QLQing Huo Liu
Qing Huo Liu
For designing or calibrating a triaxial induction well-logging tool, efficient simulations of moving the tool in an arbitrary borehole in an invaded layered medium are essential. The computational time of conventional methods is directly proportional to the number of tool positions. In this work, a domain decomposition method that utilizes the layered medium Green’s functions and loop-tree basis functions is proposed; it separates stationary subdomains from the variable subdomain where the tool moves. Efforts are saved in calculating the matrix related to the stationary subdomains. A special stationary subdomain is the boundary surface of the finite-element volume. The combination of the finite element method and the layered medium surface integral equation method minimizes the number of unknowns. Loop-tree basis functions are used to discretize the equivalent current densities on the boundary to improve numerical accuracy at low frequencies. The accuracy and efficiency of this new method are validated in well-logging simulations with a prototype induction tool including the solenoids, mandrel and sleeve in a deviated borehole and layered media.